Surge history of Múlajökull, Iceland, since 1945 detected with remote sensing data

Transcription

1 Surge history of Múlajökull, Iceland, since 1945 detected with remote sensing data Magnús Freyr Sigurkarlsson Faculty of Earth Sciences University of Iceland 2015

2

3 Surge history of Múlajökull, Iceland, since 1945 detected with remote sensing data Magnús Freyr Sigurkarlsson 10 ECTS thesis submitted in partial fulfilment of Baccalaureus Scientiarum degree in Geology Advisor Ívar Örn Benediktsson Faculty of Earth Sciences School of Engineering and Natural Sciences University of Iceland Reykjavik, February 2015

4 Surge history of Múlajökull, Iceland, since 1945 detected with remote sensing data Surges of Múlajökull since ECTS thesis submitted in partial fulfilment of a Baccalaureus Scientiarum degree in Geology Copyright 2015 Magnús Freyr Sigurkarlsson All rights reserved Faculty of Earth Sciences School of Engineering and Natural Sciences University of Iceland Askja, Sturlugata Reykjavík Telephone: Registration information: Magnús Freyr Sigurkarlsson, 2015, Surge history of Múlajökull, Iceland, since 1945 detected with remote sensing data Bachelor s thesis, Faculty of Earth Sciences, University of Iceland. Printing: Háskólaprent Reykjavík, Iceland, February 2015

5 Abstract Remote data since 1945 to 2014 was gathered through free open source data bases for the purpose to map and measure surges of the Múlajökull outlet glacier at the Hofsjökull ice cap in central Iceland. Average surge size and frequency since 1945 is measured and calculated by mapping and comparing glacier margins and moraines from various years. Surging happens on average once every 10 years with the average extension of m and affects an area of km 2 of the forefield. Múlajökull has retreated rapidly since 1995 with the largest continuous retreat during of about 600 m and has not been smaller for at least 70 years. Retreat rate of m/a remains similar throughout the period, and indicates that negative mass balance of Hofsjökull is the main reason for successively smaller surges and continuous glacier retreat in the past years. Due to smaller and less extensive surges, older moraines are preserved in the forefield and can therefore be used to reconstruct former ice-margins and surge limits. Comparison between remote sensing and end-moraine mapping in this study and ground measurements by the Icelandic Glaciological Society (IGS) suggests that at least 6 surges have taken place since This comparison also shows a significant difference between the two methods. The difference is caused by a residual error of ±15 m deriving from the difficulties with registering and distributing ground control points for georeferencing of aerial photographs. Further actions towards registration of aerial photographs and correlation between ground measurements and remote sensing measurements is needed to achieve better comparison for these two methods. Útdráttur Markmið rannsóknarinnar var að kortleggja og mæla framhlaup í Múlajökli, sem er framhlaupsjökull í sunnanverðum Hofsjökli. Út frá loftmyndum og gervitunglamyndum frá 1945 til 2014 voru jökuljaðar og jökulgarðar Múlajökuls kortlagðir og bornir saman til að meta lengd og stærð framhlaupa sem verða að jafnaði á 10 ára fresti. Við meðal-framhlaup hleypur jökullinn fram um m yfir km 2 stórt svæði. Múlajökull hefur hörfað hratt síðan 1995, um m/ári, og hefur ekki verið minni í að minnsta kosti 70 ár. Langvarandi neikvæð afkoma jökulsins virðist vera megin ástæða fyrir minni framhlaupum og hraðari hörfun. Sífellt minni framhlaup eru ástæða þess að jökulgarðar eldri framhlaupa eru enn til staðar. Kort sem sýnir framhlaup og jökulgarða frá er kynnt í lok ritgerðarinnar. Samanburður mynda frá mismunandi tímum og þróun jökulgarða leiðir í ljós að sex framhlaup hafi orðið síðan Samanburður á mælingum fjarkönnunar við svæðismælingar Jöklarannsóknafélags Íslands sýnir talsverðan mun og skoðað er hvernig stærð pixla og myndbrenglun loftmynda getur haft áhirf á mælingar, en mismunur vegna myndbrenglunar loftmynda er miðaður við ±15 m í þessari rannsókn.

6

7 Table of Contents List of Figures... vi List of tables... vii Acknowledgements... ix 1 Introduction Overview of surging glaciers Study area Methods and materials Remote sensing data background Image comparions and analysis of the Múlajökull surges Discussions Comparing remote sensing observations to ground based measurements Glacier behaviour history Surges End moraines Seasonal moraines End moraines found in the field today Limitations of this study Summary and Conclusions References Appendix A: List of all measurements done by the Icelandic Glaciology Society v

8 1 Figures Figure 1. Hillshade model of Iceland showing the study area with a close up of Múlajökull, since captured by Snævarr Guðmundsson 3 Figure 2. Glacier snout variation since 1932 (data from the Icelandic Glaciological society)...7 Figure 3. An aerial photo taken by the USA army from 1945 compared to an aerial photograph taken by the LMÍ from 1957; showing change from Figure 4. Aerial images taken by the national land survey of Iceland in 1985; showing change from Figure 5. Aerial photograph taken by the national land survey of Iceland in 1986; showing change from Figure 6. Satellite image captured in August 1990 by Landsat-5; showing change from Figure 7. Photo from 2008 showing the formation of a glacier moraine infront of Múlajökull. Photo credit by Ívar Örn Benediktsson...14 Figure 8. LiDAR image with 0.5 m resolution captured in 2008 by the Icelandic Meteorological Office and the Institute of Earth Science of the University of Iceland; showing change from Figure 9. Landsat-8 image from 2014 with 15 m resolution; showing change from Figure 10. Hydrology difference in the forefield of Múlajökull demonstrated between orthorectified aerial photograph since 1995 and a Landsat-7 satellite image from Figure 11. Locations where IGS does their measurements correlated to variation lines Figure 12. LiDAR image with 0.5 m resolution taken by the Icelandic Meteorological Office and the Institute of Earth Science of the University of Iceland in Showing all recorded surges since 1945 and moraines 25 vi

9 Tables Table 1. Overview of presented images. 6 Table 2. Remote sense data compared to ground measurements (SW) since 1945 from IGS...20 Table 3.Remote sense data compared to ground measurements (NE) since 1945 from IGS...20 Table 4. Remote sense data compared to ground measurments (center) since 1994 from IGS...21 Table 5. Remote sensing data compared to ground based measurements for surges ( )..22 vii

10 viii

11 Acknowledgements I would like to thank my supervisor, Dr. Ívar Örn Benediktsson, for professional guidance and support, and for providing maps, aerial photographs and the LiDAR hillshade used in this study. Thanks are also due to Egill E. Hákonarsson for constructive comments and useful reviews on the paper. I thank my family and friends for being there and supporting me in all my endeavors. ix

12

13 1 Introduction Some glaciers classify as so-called surge-type glaciers. They have a cyclic behavior with two phases, the active/surge phase and the quiescent phase. In a surge phase, a glacier periodically advances with one to two orders of magnitudes faster flow rate than it experiences during a quiescent phase (Benn & Evans, 2010). The distribution of surge-type glaciers is poorly understood and regional studies show different results. While longer valley glaciers tend to surge more often in some areas, piedmont valley glaciers are more likely to surge in others. Having a bed of easily eroded material underneath the glacier seems to be the only connection (Harrison & Post, 2003). Understanding fast ice flow and glacier fluctuations is becoming increasingly important because former and contemporary ice sheets are largely regulated by fast-flowing ice streams (Evans & Rea, 2003). Therefore, understanding surging glaciers may aid in the understanding of present and past larger glacial systems (Kjær et al., 2006). Surge type glaciers geomorphologically impact their forefields and leave behind landforms like crevasse ridges, flutes, drumlins, concertina eskers, hummocky moraines and glaciotectonic end moraines that can be used to reconstruct surges histories and the processes at work during surges (Evans & Rea, 2003; Kjær et al., 2008; Johnson et al., 2010; Jónsson et al., 2014). By investigating a series of aerial photographs and satellite images, the ice margin of the Múlajökull glacier, central Iceland, has been mapped with the aim of detecting surges since 1945, comparing ice-marginal positions with mapped end moraines, as well as mapping the length and aerial extent of the surges. Furthermore, the results are compared with ground-based measurements done by the Icelandic Glaciological Society (IGS) since Overview of surging glaciers Glacier surges have been explained by different theories over the last centuries. Thorarinsson (1964, 1969) rejected theories involving surges being triggered by earthquakes and Nielsen s (1937) suggestion of subglacial volcanic activities being the initiator of glacier surges. Glacier surging has been described as a cycling flow instability caused by internal forces rather than external climatic forces (Clarke et al., 1984; Raymond, 1987). William D. Harrison (2003) concludes that the main issue about the flow of glaciers is connected to the internal hydraulic system which can control pressure in the till. On the other hand till affects the subglacial hydraulic systems, so there is a fine balance between amount of water and till material. A study on the till bed of Brúarjökull in Iceland concludes that the basal motion can also occur at the sediment/bedrock interface rather than at the ice/sediment interface (Kjær et al., 2006). As of now we divide glacier surges up into two phases to describe the cycling phenomena that make up a surge. They consist of a phase between surges called the quiescent phase and an active surge phase. The quiescent phase initiates as the surging phase ends and it is the time between surges which involves glacier build up in the reservoir zone and withdrawal of the glacier (Meier & Post, 1969; Raymond, 1987). Quiescent phase is different between glacier systems, the time it usually takes for ice to build up and get ready 1

14 to surge can last from ~10 and up to ~100 years, with some examples of surges not happening again within the glacier system. The quiescent time range is somewhat related to the rate of accumulation in the accumulation zone, so if the climate in the area changes the quiescent period will likely change as well (Raymond, 1987). In Iceland surges occur irregularly which suggests that climate is not the only control (Björnsson et al., 2003; Benn and Evans, 2010). When the buildup in the reservoir zone reaches its limits ice begins to creep down thus starting the active phase. Budd (1975) suggested that a surging glacier had too little mass to sustain fast ice flow and too much mass to discharge ice using only slow flow, so in a cycling phenomena s glacier surge happens (Evans & Rea, 2003). There are developments on the glacier that can be observed and predictions made whether a surge is going to happen or not, e.g. when ice velocity is insufficient the glacier becomes stagnant or retreats causing steepening of the glacier. Elevation and velocity measurements reveal that over few months the active surge area on the glacier gets thicker and creates a bulge on the glacier. When this bulge starts to propagate forward it moves much faster than the rest of the glacier (Björnsson et al., 2003). As surging happens, the upper accumulation zone gets thinner and the glacier front advances reaching up to a thousand times faster velocity than during the quiescent phase (Evans & Rea, 2003). A glacier surge can take months and up to many years. How long a surge takes is connected to how fast ice is moving during the active surging phase (Benn & Evans, 2010). While there are examples of glacier margins not moving forward during a surge phase, it is more likely that they move at least some meters, even up to 10 km, but most commonly reaching few kilometers (Helgi Björnsson, 2003). During these ice flows, large quantities of ice move fast and apply a great amount of stress on pre-existing sediments in the field. Landforms resulting from these stresses are similar in every glacier surge forefield and define a surging glacier landsystem (Evans & Rea, 2003; Schomacker et al., 2014). Glaciotectonic end moraines, flutes and drumlins have all been linked to glacier surging. The forefield of Múlajökull is littered with end moraines and other landforms associated with surges (Johnson et al., 2010; Jónsson et al., 2014). Mapping these landforms provides a closer look on surging behavior and how to identify individual surges. In this thesis, the end moraines at Múlajökull have been mapped and correlated to surge advances detected with remote sensing data. 1.2 Study area Múlajökull is a 70 km 2, surge-type piedmont glacier of the warm-based Hofsjökull ice cap in central Iceland (figure 1). Hofsjökull rests on the largest central volcano in Iceland with a large accumulation zone in the caldera (Björnsson, 2009; Sigurðsson & Williams, 2008; Jónsson et al., 2014). It is currently losing mass at the rate m/a and is predicted to disappear in 200 years if the present climate conditions remain stable (Aðalgeirsdóttir et al., 2006). Múlajökull drains a part of the ice filled caldera through a roughly 2 km wide valley between Hjartafell (to the west) and Kerfjall (to the east) onto a relatively flat forefield where it creates a 4 km wide piedmont lobe with a ~6 km long ice margin at around 620 m above sea level (figure 1.b) (Jónsson et al., 2014). The glacier has a history of at least 4 surges since 1924 with a surge happening every years causing an ice margin advance of m. The present ice margin is situated 2 km within the outermost Little Ice Age end moraine and since the last major surge in 1992 it has retreated 700 m (Johnson et al., 2010; Jónsson et al., 2014). An active drumlin field has been exposed since the glacier started to retreat rapidly in the last century. This unique field is of 2

15 great interest to scientists for it is the only active drumlin field known today (Johnson et al., 2010). Hjartfell Kerfjall (b) Figure 1. (a) Hillshade model of Iceland, black box represents the location of Múlajökull. (b) Overview photo of Múlajökull from captured by Snævarr Guðmundsson. Taken from: suður, Snævarr Guðmundsson,

16

17 2 Methods and materials Series of aerial photographs were gathered from the National Land Survey of Iceland (Landmælingar Íslands; LMÍ) which provides an open free source aerial photograph collection. Satellite images for missing years were gathered from USGS EarthExplorer which provides free Landsat products. It should be noted that satellite images have much lower resolution than aerial photographs. Each pixel in a satellite image is usually 30x30 m while the aerial photographs usually have a pixel resolution of 1x1 m or lower (table 1). Ground control points on the images were marked and georeferenced to an already orthorectified image from 1995 in the UTM/WGS84 coordinate system. Residual error is the difference in distance between the real position in the area and the georeferenced aerial photo. Any one pixel on the georeferenced aerial photo (for aerial photos used in the thesis) has the maximum difference of 15 m from the real position and minimum difference of 0.4 m from the real position. The glacier margins were mapped from aerial photographs (presented as lines along the margins on maps) for years: 1945, 1957, 1960, 1984, 1985, 1986 and 1995, the margins were also mapped from satellite images for years: 1973, 1987, 1988, 1990, 1991, 1994, 2000, 2008, 2009, 2010 and A LiDAR image with 0.5 m resolution recorded by the Icelandic Meteorological Office and the Institute of Earth Sciences. University of Iceland, in 2008 was also used for mapping and as a basemap for recent years (Jóhannesson et al., 2013; Jónsson et al., 2014). Ten survey lines to measure distances between margins were chosen with approximately 775 m between them to acquire even distribution. Glacier variations were mapped with polygons between years: , , , and Polygon areas were calculated to give an approximate of aerial change, measured in km 2. Yellow polygons represent glacier advance and red polygons represent glacier retreat. All data was processed using ESRI ArcGIS Remote sensing data background In July 1972 the first Landsat satellite was launched called Landsat-1. Landsat one took photos in four spectral bands, visible spectrum and the near-infrared electromagnetic spectrum, providing images with about 80 m pixel resolution. Landsat-5 was then launched in 1984 as well as Landsat-7 in 1999 and Landsat-8 in 2013 all of them equipped with a TM sensor which provided ~28.5 m resolution (table.1). However band 8 in Landsat-8 provides panchromatic images with a 15 m resolution. With the right tools in ArcMap 10.1 it is possible to apply this resolution to visible spectrum bands, giving true color satellite images 15 m resolution (Hall et al., 2003). Glacier margin retreat/advance can be determined if the retreat/advance is larger than the pixel size of the image (Williams et al., 1997). All aerial photographs have an uncertainty of 1 m or lower because of pixel size. Residual error affects accuracy for aerial photographs and in this case the error is considered ±15 m. 5

18 Table 1. Overview of presented images. Year Type Origin Resolution (m) 1945 Aerial photograph US Military Aerial Photograph LMÍ Aerial Photograph LMÍ Satellite Image Landsat LiDAR Image Optech ALTM 3100 laser scanner Satellite Image Landsat Image comparison and analysis of the Múlajökull surges Repeated measurements or multiple researches on a single surging glacier are rare (Björnsson et al., 2003). Although the IGS started their measurements on glacier fronts back in the 1930s their measurements only covered a very small area of the Múlajökull forefield; hence, they do not give a complete picture of glacier front changes (Björnsson et al., 2003). Observations with remote sensing data give us a bigger picture of the changes that Múlajökull has gone through since Short indication before a surge makes it hard to monitor the activity on the ground, thus making remote sense data a viable option. Studying glacier change from space is not a new approach. However, papers comparing historical aerial photographs with ground based measurements over a long period are uncommon (Hall et al., 2003). Furthermore no references were found on the comparison of long time ground observation and historical aerial photographs of surge-type glaciers From this period two images will be compared. An aerial photograph from 1945 is the earliest image for comparison (table 1). This image displays Múlajökull during a quiescent phase. A graph showing the extent of the glacier margin over years was compiled from the IGS database and presented in a paper (Benediktsson et al., in review) (figure 2). The data from IGS show that the glacier was slowly retreating from the 1920s until 1954 when a surge took place. During the 1954 surge the glacier advanced about 200 m (figure 2). An aerial photograph taken by LMÍ in 1957 shows the ice margin further out in the forefield than it was in On the 1957 aerial photograph a moraine in front of the glacier margin is observed (figure 3). Between the 1957 ice margin and the 1954 end moraine, there are 6

19 two to three visible smaller moraines. These are most likely seasonal moraines that represent the stepwise annual recession of the ice margin from 1954 to 1957 (e.g. Krüger 1995; Johnson et al., 2010; Jónsson et al., 2014). The average length difference between the 1945 ice margin and the end moraine in front of the 1957 margin is measured to be 80.3 m. Line 5 reveals the largest extension of 161 m (figure 3). The greatest advance occurred in the central forefield whilst the advance at the NE and SW sides was considerably smaller. It is worth noting that between 1945 and 1954, the ice margin probably continued to retreat as part of the quiescent phase behavior. Therefore, the surge advance was most likely somewhat larger than indicated by the numbers above, which, as a result, can be regarded as minimum numbers. The comparison between the 1945 and 1954 images also shows that the NW and SW edges of the Múlajökull lobe are not affected by the surge and, in fact, retreat during this period. Decrease in lagoons and water outlets are observed between 1945 and 1957 images. Where the glacier is further out during 1957 it covers a larger area and overruns parts of the bigger lagoons observed on the 1945 image. Calculated increase in glaciated area extent from 1945 to 1954 is 0.63 m 2 (figure 3). Figure 2. Glacier front variations of Múlajökull since 1932 (from Benediktsson et al., in review, based on the original data from the Icelandic Glaciological Society). 7

20 Figure 3. Comparison of the ice margins in 1945 and 1957, shown on the aerial photograph from 1945 (recorded by the USA army, see Table 1). Both photos were geometrically corrected in ArcMap The 1945 ice-marginal position is shown with a yellow line and the red line indicates the end moraine and the maximum extent of the surge in Yellow polygon shows the minimum area covered by the 1954 surge advance and red polygons show glacier retreat from 1945 to During 1957 to 1984 there are no aerial images of Múlajökull available from LMÍ. Satellite images gathered for this period were deemed unusable due to missing area coverage and large pixel size. The glacier front variations recorded by IGS (figure 2) shows three glacier advances during this period: a small one in 1966, a major advance in the early 1970 s and a minor glacier advance in IGS reports an advance by 71 m in 1966, 370 m advance in and 17 m advance in 1979 (figure 2). Even though the satellite image is barely usable because of low resolution, an increase in glacier extent can still be observed from 1960 to 1973 due to the surge. The 1966 surge is not observable with available images. 8

21 High quality aerial photographs acquired from LMÍ show Múlajökull s position during 1984, 1985 and Multiple aerial images from 1984 and 1985 were needed to create a mosaic of the whole glacier margin. While mapping margins and moraines, images were switched around so the black image borders would not affect the outcome (figure 4). From 1984 to 1985 the glacier retreats slowly in the center but is close to stationary on the NW and SW sides. Small changes in lagoon sizes are observed as they grow when the glacier retreats. This is well known from other currently retreating glaciers in Iceland (Schomacker, 2010). Figure 2 shows a retreat between years 1984 to Between the 1954 end moraine (situated about 200 m in front of the 1985 ice margin) and the 1985 ice margin, two unknown moraines are observed. Unknown moraine-1 cross-cuts the 1954 end moraine in some parts in the central forefield and unknown moraine-2 crosscuts unknown moraine-1 at the northern side but does not extend as far out at the front of the glacier (see figure 4.b). Most probably, these moraines result from the surges in and Aerial image from 1986 shows the glacier further out in the forefield than it was in Many large lagoons ( m m 2 ) that are visible on the 1985 image disappear underneath the glacier in The average surge advance from 1985 to 1986 is 165 m with measuring line 4 showing the greatest advance of 288 m (figure 4). The surge completely overrode unknown moraine-2 and partly overrode unknown moraine-1 at the northern side (figure 4). The end moraine from 1954 is not overrun by the surge advance. The surge advance from 1985 to 1986 covered an area of km 2. 9

22 Unknown moraine -1 Unknown moraine -2 Glacier moraine 1954 (b) Figure 4. Aerial images taken by the national land survey of Iceland in Positions of glacier margins from 1985 and 1986 are shown as purple and green lines. The yellow polygon shows glacier extension from 1985 to Red lines between glacier margins are used to calculate average surge length. Moraines of unknown ages (though probably and 1979) are shown with black and white lines. The 1954 surge end moraine is shown with a red line. Red box indicates the location of figure b. (b) close up of the two unknown end moraines and the end moraine from The aerial photograph from 1986 is captured around the culmination of the 1986 surge and shows the glacier relatively far out in the forefield. The end moraine from 1954 and unknown moraine-1 can still be observed on the 1986 photo in front of the glacier (Figure 5 b). From 1986 to 1991 there are no aerial photographs available from LMÍ but satellite images from Landsat-5 during this period were used instead. A close approach to the real 1991 ice-marginal position was made by combining information from satellite images from 1990 and using measurements from IGS. 10

23 Information observed from the satellite images show a net retreat during this period. Quality difference between the aerial photograph and satellite image makes it hard to observe any measurable change in water outlets or lagoon sizes. Earlier observations have concluded that with glacier retreat more water stores in front of the glacier (Schomacker, 2010). Observing changes that affect large areas is possible by using only satellite images but more detailed measurements have high uncertainty error. Mapping the margin is possible and by doing so a clear change in glacier area is observed. The area of Múlajökull decreased from 1986 to 1991 by about km 2. This is similar to the area increase during the 1986 surge, meaning that the 1991 ice margin would be in a situation similar to that before the 1986 surge. The glacier retreat is similar over the whole margin with an average retreat of 163 m. Measuring line 10 shows the longest retreat of 247 m (see figure 5). 11

24 Glacier moraine 1954 Unknown moraine -1 (b) Figure 5. Aerial photograph taken by LMÍ in Positions of glacier margins are presented as green and brown lines. Red polygon represents the glacier retreat from 1986 to Red lines between the glacier margins are used to calculate average length retreat. Red box indicates zoomed in area. (b) Close up of the 1954 moraine and unknown moraine There are no aerial photographs from this period available from LMÍ. Satellite images from Landsat-5 were gathered and used instead (table 1). As previously mentioned, only large scale measurements are valid using satellite images with high uncertainty. Both satellite images that were used for comparison were taken in august. 12

25 On the image from 1994 the glacier is further out on the forefield than in A comparison of the ice-marginal positions in 1991 and 1994 shows an average advance of 164 m along the margin. The greatest advance of 264 m is at line number 5 in the central forefield (see figure 6). For comparison, the data from IGS shows an advance of ca. 200 m (figure 2). The glacier continued to retreat over 1991, and started retreating again in therefore the glacier advance was most likely larger than the numbers above indicate. The glacier surged in out to the 1986 moraine and overrode it in most places in the central forefield. Parts of the 1986 moraine remained untouched at the NE and SW sides. Advance during also affected the 1954 glacier moraine and the unknown- 1 moraine by crosscutting them in the central forefield. The increase in glacier coverage from 1991 to 1994 was km 2. Gleacier margin 1994 Figure 6. Satellite image captured in August 1990 by Landsat-5. Positions of glacier margins for 1991 and 1994 are presented as brown and blue lines. Yellow polygon represents area extension from Red lines between the glacier margins are used to calculate average length extension from

26 In 2008 the Icelandic Meteorological Office and the Institute of Earth Sciences, University of Iceland, captured a LiDAR image of the area (Jóhannesson et al., 2013). For this period a satellite image taken by Landsat-5 from 2009 was used for comparison. There is a thin cloud cover at the northern edge of the glacier margin on the 2009 satellite image. A moraine with unknown origin appears in front of the glacier margin in 2008 (unknown moraine-3; figure 8.b). Differences in ice-marginal positions are observed when comparing the 2008 and 2009 data. An increase in glacier area by 0.44 km 2 can be observed along the glacier margin, with the greatest difference of 147 m measured along line number four (figure 8). On the 2009 image, the ice margin cross-cuts unknown moraine-3 in various places except in the northern part. The data from IGS suggest a small surge advance of only 10 m in This difference is interesting because already on the Lidar image from 2008, a prominent end moraine can be observed immediately at the ice margin, indicating that a surge had or was about to culminate in 2008 (Jónsson et al., 2014). Furthermore ground observations during field work in 2009 concluded that a moraine was forming in front of the glacier margin (figure 7). Figure 7. A prominent end moraine right at the ice margin in the central forefield in 2009, most likely just after the culmination of the surge. Person for scale. Photo credit by Ívar Örn Benediktsson. 14

27 Unknown moraine-3 (b) Figure 8. LiDAR image with 0.5m resolution captured in 2008 by the Icelandic Meteorological Office and the Institute of Earth Sciences, University of Iceland (Jóhannesson et al., 2013). Position of glacier margins and end moraines are presented with various colors as individual lines. Yellow polygons show change from 2008 to 2009 and pink polygons show retreat from 2009 to 2008 which is most likely an error caused from misinterpretation by the cloud cover. Red lines between margins are used to calculate average length extension. (b) Zoomed in area of the unknown moraine in front of the 2008 glacier margin Two images were used to capture change between 1995 and An orthophotograph with 0.5 m resolution was used to map the glacier margin in For comparison, a Landsat-8 satellite image from august 2014 was acquired (table 1). 15

28 In 1995 the glacier is far out on the forefield and just starting to retreat after a significant surge advance in On the satellite image from 2014, the glacier margin is more irregular and farther up-ice than ever before since 1945, suggesting that Múlajökull has never been smaller during the past 70 years (figure 9). Large lagoons have formed in front of the margin and new landforms such as drumlins are visible in the forefield (Johnson et al., 2010; Jónsson et al., 2014; Benediktsson et al., in review). The loss of glacier area since 1995 is around 4 km 2. Average glacier retreat along the margin is 598 m and the greatest retreat is 728 m along line 5. Progressive shortening on the sides and extension in the front is observed between 1945 margin and the 1995 margin. Glacier margin 1945 Figure 9. Landsat-8 satellite image taken in 2014 with 15 m pixel resolution. Position of glacier margins are presented with orange, yellow and dark pink lines. Red polygon represents retreat of the glacier from 1995 to Red lines between the glacier margins are used to calculate average length retreat. 16

29 Bewtween 1995 and 2014, Múlajökull lost about 4km 2 of its area. As a result, an equivalent area of the former subglacial bed (now forefield) has become exposed. The hydrology in the forefield changed from and the glacier margin became more irregular. When the 1995 orthorectified aerial image is laid over the 2014 satellite image, a significant increase in the area of proglacial and ice-marginal lakes can be observed. Blue polygons show large lagoons and water outlets in 2014 (figure 10) Figure 10. Hydrology difference in the forefield of Múlajökull demonstrated between orthorectified aerial photograph since 1995 and a Landsat-7 satellite image. Blue polygons display lagoons and rivers in

30 4 Discussions Glacier margin retreat/advance can be determined if the retreat/advance is larger than the pixel size of the image (Williams et al., 1997). According to this approach it is clear that aerial photographs are of better use when determining glacier margin changes and to obtain accurate measurements. Therefore sequences of aerial photographs are viable to determine glacier margin change. However, while glaciers are usually advancing or retreating, surgetype glaciers can advance with a short notice. For these reasons continuous sequences of aerial photographs need to be available. When the aerial photograph sequence isn t continuous, low-resolution satellite images that are free of charge are a limited source for accurate year to year measurements, compared to aerial photographs. Furthermore it is very important to know the measurement errors that follow pixel sizes and residual errors in aerial and satellite imagery. When deciding where the glacier margin terminates, local knowledge of the area is considered of critical importance and might affect some results where satellite images were used (Williams et al., 1997). However on the aerial photographs with an error of ±1 m, a local knowledge is not as important and thus, misinterpretation for where the glacier margin terminates is unlikely. 4.1 Comparing remote sensing observations to ground based measurements Annual measurements on surging glacier outlets since 1932 done by the IGS serve as a fundamental database when testing the reliability of remote sensing of the advances and retreats of surge-type glaciers (Björnsson et al., 2003). Today IGS does measurements on three different parts of the Múlajökull glacier (figure 11). Earliest data was gathered on the northern side of the ice margin where IGS started their measurements in Measurements on the western side started in 1937 and measurements for the center of the margin were initiated in However, the earliest available aerial photographs are from 1945, which defines the period of comparison to the IGS dataset. Survey lines compared to ground measurements from IGS were decided to be: line number 2 which is close to IGS NE measurements, line number 5 is close to their central measurements and line number 9 is close to their SW measurements (figure 11). Uncertainty error given when comparing years that both have aerial photographs available for comparison is considered to be ±1 m. When only one of the years has an aerial photograph available for comparison and another image is from a satellite the uncertainty error is ±15m for residual error and ± 30 m for pixel size, except for the satellite image from 2014 which has a pixel size of ±15 m. 18

31 Figure 11. Locations of the IGS survey lines shown with the measurement lines of this study (red). Insert map taken from and laid on top of the 2008 Lidar image. 19

32 Table 2. Remote sensing data (Line 9) compared to IGS s ground measurements (SW) since A negative number accounts for retreat and a positive number accounts for advance over the specific period. A positive number in the difference column means that remote sense data shows a longer glacier extension or a shorter glacier retreat compared to IGS s ground measurements. Years SW (m) Line 9 (m) Difference (m) ± ± ± ± ± ± ± ±30 10 Table 3. Remote sensing data (Line 2) compared to IGS s ground measurements (NE) since A negative number accounts for retreat and a positive number accounts for advance over the specific period. A positive number in the difference column means that remote sense data shows a longer glacier extension or a shorter glacier retreat compared to IGS s ground measurements. Years NE (m) Line 2 (m) Difference (m) ± ± ± ± ± ± ± ±

33 Table 4. Remote sense data (Line 5) compared to IGS s ground measurements (Central) since A negative number accounts for retreat and a positive number accounts for advance over the specific period. A positive number in the difference column means that remote sense data shows a longer glacier extension or less glacier retreat compared to IGS s ground measurements. Years Central (m) Line 5 (m) Difference (m) ± ± History of glacier behavior Numbers for the southwest side show a net retreat in the area, numbers agree well with remote sensing data except for (table 2). Ice margin line for 1991 was built with the help of a satellite image from 1990 and measurements from IGS for retreat during 1991; this could result in some misinterpretation for the 1991 ice margin and the uncertainty error should perhaps be larger. During quiescent phase ( , , ), Múlajökull shows a retreat rate of about m/a on the SW side. However from there were at least two surges and possibly three (1966, and 1979) (Björnsson et al., 2003). As a result the net retreat over this period is relatively small with average retreat of about 1 m/a. IGS measurements during quiescent phases for Múlajökul suggest the average retreat rate since 1945 on the SW side is around m/a. During the period from the SW side of Múlajökull retreats even though there were multiply glacier advances during the period. Since 1995 there is a record of one small surge that slows down the net retreat. However, the net retreat from is the largest retreat observed since 1945 and according to data from IGS it is the greatest net retreat since at least 1937 (appendix A). Net retreat from is about 30 m/a, which is similar to earlier retreat rates during quiescent phase. This shows that during quiescent phase the glacier has shown similar retreat rates since The reason for a large retreat in the last years could therefore be connected to the lack of ice in the reservoir zone during surging phase rather than the rate of retreat. Recent studies on the thinning of the Hofsjökull ice cap (Jóhannesson et al., 2013) shows that during the glacier margin on all sides is getting thinner with average rate of 5 m/a, and around 800 m above sea level the maximum net lowering of ice since 1986 exceeds 80 m on the eastern side of Hofsjökull, suggesting that the surge accumulation zone is getting less ice. Greater advance/retreat fluctuations on the NE side (table 3) could be affected by a larger supply of subglacial meltwater in that area, as indicated by the Arnarfellskvísl River emerging from underneath the glacier. Earlier models for ice flow mechanism have shown 21

34 that deformation of water saturated sediments sustain fast ice flow (Engelhardt & Kamp, 1998). As a result, Múlajökull may experience greater advances at the NE side than on the SW side due to more subglacial meltwater. From remote observations it also seems that larger lagoons accumulate in the NE area than in the SW area, especially during the last years (figure 10). Since 2000, glacier lagoons and lakes have been expanding and multiple valley glaciers have developed calving fronts as a response to climate change (Schomacker, 2010). This tells us that more meltwater is coming from the glacier suggesting that during the last years there is a faster net outflow from Múlajökull than it was during /2000. Change in the forefield was extensive from Hydrology in the field changes significantly when the glacier retreated to its new position in In this newly exposed area, multiple glacial landforms, such as outwash plains, flutes and crevasse-fill ridges, were mapped by Jónsson et al. (2014). Drumlin field in front of the glacier margin is also well observable on the 2008 LiDAR image (figure 12). Recent papers have identified three zones in the forefield of Múlajökull that represent different morphologies (Jónsson et al., 2014). Ground measurements suggest that the greatest glacier retreat is on the NE side and least retreat at the SW side. However remote sense data shows that the fastest glacier retreat is at the central of the margin (table 4) and gradually slows down to both sides although it slows down less towards NE (figure 9) Surges According to Björnsson et al., 2003 there are recorded surges in Múlajökull during years: 1924, 1954, 1966, 1971, 1979, 1986, 1992 and possibly in It was possible to observe and do measurements for surges in four time periods by using aerial photographs and satellite images (table 5). The images above cover glacier advances in 1954, 1986, 1992 and According to remote sensing data the glacier advances m into the forefield and affects a foreland area of km 2. Table 5. Remote sensing data (Lines 9, 2 and lines 4, 5) compared to IGS s ground measurements for glacier advances ( ). A minus number accounts for retreat and a positive number accounts for advance over the specific period. A positive number in the difference column means that remote sense data shows a longer glacier extension or less glacier retreat compared to IGS s ground measurements Observed advance IGS data for SW/NE (m) Remote sensed data for line 9 and 2 (m) Difference (m) Maximum advance either line 4 or 5 (m) Area affected (km 2 ) /104 80/3±15 32/ /318 51/162±15 17/ / /179±30 79/ /10 106/36±30 125/

35 Some difficulties are with this comparison; surge advance varies a lot along the margin and locations compared to are not exactly the same. Residual error for aerial photographs could be playing a more significant role than previously thought, and satellite derived measurements are less conclusive because uncertainty error resulting from pixel sizes. The glacier margin during recession becomes abundant with debris, seen from a satellite image the margin might look like surrounding areas thus making it harder to determine its exact location (Williams et al., 1997) End moraines End moraines are formed at the margins of advancing or stationary glaciers (Benn & Evans, 2010). Therefore, the end moraines can be postulated as delineating the maximum extent of a glacier advance (Benediktsson et al., 2009). Three end moraines with unknown ages were observed. Two were created sometime in and one sometime in On the aerial photograph from 1985, unknown moraine-2 was observed to cross-cut unknown moraine-1. This observation tells us that unknown moraine-2 is younger than unknown moraine-1. There are records of surges happening in 1966, and According to numbers from IGS, the glacier retreated 159 m during Distance measured on images above, between unknown moraine-2 and the 1985 glacier margin varies along the margin from about m. Therefore, it is likely that unknown moraine-2 is the end moraine created in Furthermore, distance during according to IGS is a retreat of 333 m. Distance from unknown moraine-1 to the 1985 glacier margin varies along the margin from m. Therefore, it is concluded that unknown moraine-1 is the moraine formed by the surge in A table covering all documented surges in Iceland was compiled by Björnsson et al. (2003). According to the table of surges, 6-7 surges have occurred in Múlajökull since A question mark is put on a surge in 2002, records from IGS are said to be faulted for Unknown moraine-3 was observed right in front of the glacier margin on the LiDAR image from 2008, so it is formed sometime in the period from During field work in the summer of 2009 a moraine right in front of the glacier margin looked very fresh (figure 7). Furthermore the glacier was slumping on top of the margin on various spots along the margin in 2009 so it was partly formed in 2008 and continued to form in 2009 (Ívar Örn Benediktsson, personal communication, January 22, 2015). Unknown moraine-3 is therefore considered as the moraine formed in Seasonal moraines Smaller moraines were observed between the 1954 push moraine and 1957 glacier margin that are interpreted as annual winter-advances between those years (Krüger, 1995). Small moraines were found in field studies during between the 1992 and 2008 end moraines, varying from 8 to 15 and interpreted as annual winter-advances (Jónsson et al., 2014). These findings suggest that there was not a surge in 2002 or

36 4.1.5 End moraines found in the field today A period of colder climate which affected climate in Europe strongly is called the Little Ice Age (LIA) and lasted from the 16 th -mid 19 th century (Mann et al., 2002). During the LIA, ice caps expanded and many outlet glaciers advanced far out into their forefields (Björnsson & Pálsson, 2008). The outermost end moraine in the forefield of Múlajökull is most likely formed during LIA and is locally known as Arnarfellsmúlar (Jónsson et al., 2014). Between Arnarfellsmúlar and the 1992 end moraine there are at least five other end moraines (Jónsson et al., 2014). The extent of surge advances during is very similar, resulting in the formation of multiple end moraines around the same position (figure 12). Progressive change over this period suggests a decreasing extent to the NE and SW sides preventing younger glacier advances to override older moraines. At the center there is more cross-cutting between moraines resulting in distributed parts of moraines with different ages. The 1992 moraine is extensive and continuous across the forefield but parts of older moraines are found in front of the 1992 margin and large parts are continuous along the NE and SW sides. During the glacier retreats more rapidly than it did during As a result a large area in the forefield is exposed and the glacier seems to have retreated to a new surge zone, with the end moraine from being the footprint of the first surge in this new area. 24

37 Figure 12. End moraines formed between 1945 and 2008 presented with various colors on the basis of a LiDAR image with 0.5 m resolution (Jóhannesson et al., 2013). Thicker lines show what parts of the moraines are still visible in the field. Thinner lines show the maximum surge extent. The 1979 moraine is overridden and not visible anymore. 25

38 4.2 Limitations of this study Measurement qualities are primarily affected by pixel sizes and residual errors. Satellite images have a higher uncertainty when it comes to pixel sizes but no residual error for they are all set in the same coordinate system. Determining the glacier margin on satellite images is however an issue whereas the glacier retreats more sediment is deposited in the front and often resulting in giving the glacier margin a very similar reflectance as its surrounding area. Residual error is considered to be ±15 m for this study, mapped ice margins from aerial photographs seemed to reflect well to the satellite images and the already orthorectified image from Furthermore where multiple photographs were needed to build a mosaic of the margin the photographs fitted to each other very well giving the georeferencing work some reliability. Highest uncertainty applies to comparison for and where pixel size effect is ±30 m and residual error ±15 m. Remote sense data does not seem to compare well to the ground measurements. There are some quality factors within the ground based measurements as well e.g. ground measurements are missing years: 1961, 1963, 1968, 1976, 1980, 2001, 2002 and 2005 (appendix A). Missing years are due to unsatisfied measurements or it was decided to skip those years due to other circumstances. For some measurements there is also added that they are not very precise because of obstacles in the field (lagoons) Summary and conclusions Aerial photographs give a good overview of the glacier history; surges can be observed and dated. The history of geomorphological development in the forefield can be tracked and used to determine amount of surges. However, no trace of a surge in 1966 was found as an end moraine from that surge was most likely overridden by the surge. Precise measurements are possible with aerial photographs, with that said; the registering process for aerial photography is likely the most important factor to obtain accurate measurements. This process can be hard for glaciated areas for two reasons. First; the area is constantly changing. Second; mapping glacier change usually involves a large portion of the image to be glaciated, registering areas on a glaciated surface is hardly accurate so a residual error for poorly distributed registration points is likely. From observing and mapping the history of Múlajökull and its forefield following conclusions are drawn: There have been at least 6 surges since 1945, this is gathered from observations involving surges during: , , and looking at the development of moraines in the forefield; two created during concluded to be from the and 1979 surges, and one created during concluded to represent a surge in Surges occur on average at least once every 10 years, usually advancing m into the forefield and increasing glaciated area by km 2. Larger surges happened in: , 1986 and 1992, advancing about 300 m and covering an area of 1 km 2. Since 1945, the extent of surges has become successively smaller at NE