080

20/3/2015

1

EVALUATING CHANGES TO ARCTIC COASTAL BLUFFS USING REPEAT AERIAL PHOTOGRAPHY AND STRUCTUREFROM-MOTION ELEVATION MODELS ANN E. GIBBS1, MATT NOLAN2, BRUCE M. RICHMOND1, 1. 2.

U.S. Geological Survey, Pacific Coastal and Marine Science Center, 400 Natural Bridges Drive, Santa Cruz, CA 95060. [email protected], [email protected] Institute of Northern Engineering, University of Alaska Fairbanks, Fairbanks, AK 99775, [email protected]

Abstract: Aerial photography was collected over Barter Island, Alaska, USA in July and September, 2014. Digital elevation models (DEMs) were derived from the photography using Structure-from-Motion (SfM) techniques, resulting in orthophotomosaics with pixel resolutions of 0.18 and 0.11 cm and DEMs with pixel resolutions of 0.18 and 0.22 cm, for each survey, respectively. Comparison of the image and elevation data along a 2.7 km long section of coastal bluffs shows that considerable bluff retreat and morphological change occurred in just over 2 months, including up to 5 meters of retreat of the top edge of the bluffs and a cumulative loss of nearly 28,000 m3 ± 540 m3 of material, primarily through the removal of debris at the base of the bluffs.

Introduction Chronic and widespread coastal erosion along the northern coast of Alaska is threatening traditional lifestyles, sensitive ecosystems, energy and defense related infrastructure, and large tracts of Alaskan Native, State, and Federally managed land. A recently completed shoreline change analysis along the Alaskan Beaufort Sea coast, based on historical NOAA topographic sheets and aerial and satellite photography, indicates shoreline change rates average -1.7 m/yr and range between -18.6 m/yr and +10.9 m/yr between 1947 and 2012 (Gibbs and Richmond, 2015). The historical analysis provides important information regarding long-term rates of change but does not convey details on the timing or processes driving it. Aerial- and ground-based photography can provide valuable information about the physical environment in space and time, including the presence or absence of shorefast ice or snow, beach characteristics including texture, morphology (for example erosional scarps and ice-push ridges), wrack lines produced during storm surge events, bluff failure mechanisms, and habitat identification. Recent advances in digital photogrammetry applied to aerial photography can also be

080

20/3/2015

2

used to construct high quality Digital Elevation Models (DEMs) at a relatively low cost. Repeat aerial surveys and associated DEM construction serve as a powerful monitoring tool that can be used to quantify linear and volumetric change, and, if conducted frequently enough, provide insights into the mechanisms responsible for coastal change. The village of Kaktovik and adjacent U.S. Air Force radar site on Barter Island, Alaska, USA are fronted by an eroding coastal bluff where attempts to control bluff erosion with shore protection structures were undertaken more than a decade ago. Here we evaluate seasonal changes in shoreline and coastal bluff position and morphology using vertical aerial photography and DEMs acquired in early (1-July) and late (7-Sep) summer, 2014. Study Location Barter Island, considered the gateway to the Arctic National Wildlife Reserve (ANWR), is located on the northeast coast of Alaska approximately 120 km west of the U.S.-Canadian border on the Alaskan Beaufort Sea coast (Fig. 1). The island has a long history of Inupiat occupation and was a major trade center until the late 19th century (State of Alaska, 2015). The City of Kaktovik, on the north shore of Barter Island, was incorporated in 1971 and had a population of 239 as of the 2010 census. Similar to other Native Villages in northern Alaska, subsistence hunting, fishing and whaling play a major role in the local economy. In the 1950s, the U.S. Air Force built an airstrip and Defense Early Warning Line (DEW) radar station on the island. The DEW site was deactivated in 1989 and in 1990 upgraded to part of the North Warning System Long Range Radar Site (LRRS) which continues operations today. Access to the island is limited to barges and aircraft, which currently land on the gravel/sand spit to the east of Kaktovik. The existing (2015) airstrip, constructed on the low-lying spit, is less than 2 m above sea level and floods nearly every summer during periods of elevated ocean water-levels. A new airport is under construction on the relatively higher tundra on the southern part of the island (Fig. 1). Coastal bluffs front the elevated tundra core of the island and range in height from a few meters to more than ten meters high. Barter Island, like the entire north slope of Alaska, is in a zone of continuous permafrost, where the ground is frozen throughout the year. Only a shallow active layer (generally less than 1 m deep) of surface material is subject to thawing during the summer season. The coastal permafrost bluffs are poorly consolidated and consist of a complex sequence of material ranging from dense marine clay at the base, sands and gravel thought to be of fluvial origin, units of massive sand of unknown origin, massive ice which

080

20/3/2015

3

has been interpreted as buried glacial ice (Jorgenson and Shur, 2008), wedge ice, thermokarst cave ice, aeolian silts and sands, and a surface peat layer. Aerial lidar DEMs obtained in 2009 (U.S. Geological Survey, 2015) revealed higher bluff elevations across the central portion of the island where field observations of bluff stratigraphy showed multi-layered stratification. The lower elevation outer flanks of the exposed bluff face consist of homogenous layers of sandy-silt below the surface peat layer (Gibbs and others, 2010). The island is flanked by broad, lowlying (< 2m high) sand and gravel spits extending both east and west from the topographically higher tundra core of the island. Long-term coastal bluff retreat rates on Barter Island averaged 1.3 m/yr between 1955 and 2010, however, local retreat of up to 20 m in a single year has been observed (for example in 2008; Gibbs and others, 2010). Mechanisms for bluff failure along this coast are poorly understood but likely include a combination of thermal and mechanical processes. Recession rates appear to be largely dependent on ice content, the frequency and intensity of storms, run-up elevation, and seawater and air temperatures. Methods Image Acquisition and DEM Construction Aerial photogrammetry, a technique where aerial photographs are turned into topographic maps and image mosaics, has long been used as a tool to map coastal change (see for example Dolan and others, 1978; McBride, 1989; Moore, 2000). Recent advances in digital photogrammetric technology, including improvements to consumer-grade cameras, GPS processing techniques, desktop computer processing capabilities, and the development of powerful photogrammetric software using Structure-from-Motion (SfM) algorithms (Koenderink and Van Doorn, 1991; Westoby and others, 2012; Nolan and others, 2015), have revolutionized the acquisition of low-cost digital imagery and elevation data. SfM software, as well as traditional photogrammetric-processing software, triangulate the positions of points on the ground that have been imaged multiple times in overlapping photographs to create a point cloud of X, Y, and Z values defining the measured surface. This point cloud can then be gridded into a digital elevation model (DEM) or an orthometrically corrected image mosaic. Numerous hardware and software packages and imaging platforms can be used to create the DEMs and orthophotgraphs. Here we use Fodar™ (http://fairbanksfodar.com), a proprietary mapping system that incorporates a Nikon D800E digital single lens

080

20/3/2015

4

reflex camera and dual-frequency GPS mounted in a Cessna 170 aircraft. Imagery was processed using Agisoft Photoscan software to create orthophotographs and DEMs following methodologies outlined in Nolan and others (2015) and at http://fairbanksfodar.com. Aerial photography was acquired in early (1-July) and late summer (7-Sept) of 2014. The surveys were opportunistic and reconnaissance in design and not optimized for maximum resolution and precision. Flight lines and altitudes differed between the two surveys, as did the resulting resolution of the orthoimages and DEMs (Table 1). The September orthoimage and DEM were adjusted to the WGS84(G1674) geographic coordinate system and ellipsoid using three ground control points (GCP) surveyed using dual-frequency GPS equipment in September 2014. The July data were adjusted using image-to-image and DEMto-DEM correlation with the September data.

Fig. 1 Lidar derived elevation map of Barter Island showing locations described in the text.

Feature Extraction and Change Analysis Two and 3-dimesional analyses of the data were completed for a 2.7 km stretch of the bluffed coast (fig. 1). Two-dimensional features, including the position of the bluff edge (top and base, typically recognized as the seaward edge of the debris fan) and the shoreline (based on the land-water interface), were manually delineated in a GIS using the orthoimage and elevation and slope rasters. The base of the bluff was typically recognized as the seaward edge of the debris fan, however, in areas where snow or ice was present in the July data, the base of the bluff was delineated as the landward boundary between the ice/snow and

080

20/3/2015

5

exposed debris fan. Change in the position of the features was calculated every 10 m along the 2.7 km stretch of coast using the Digital Shoreline Analysis System (Theiler and others, 2009). The two DEM surfaces were compared to determine sediment volume change of the coastal bluffs. The July surface was subtracted from the September surface in a GIS and the resulting surface was then clipped to an ‘active bluff ’ polygon. The active bluff was defined on the seaward side by the base of the bluff (typically from the July surface) and the landwardmost top edge of bluff, which in this case was always from the September survey. The resulting difference surface was multiplied by the raster pixel resolution (0.23 x 0.23 cm) to produce volume change. Table 1. Data acquisition and adjustment parameters Date Flown

Altitude

Resolution

GCP Comparison

Image

DEM

Horizontal

Vertical

Jul 1, 2014

740 m

18 cm

18 cm

±8 cm

7 cm

Sep 9, 2014

450 m

11 cm

22 cm

±5 cm

7 cm

Accuracy and Error Analysis Robust accuracy and precision assessments based on surveyed ground control and temporally-consistent repeat surveys were not completed for this study due to the reconnaissance nature of the data collection and limited coincident GPS groundtruthing surveys (2 points). Extensive testing of the accuracy, precision and noise level of the Fodar™ method of photogrammetry has been completed, however, as part of a study mapping snow depths from manned aircraft (Nolan and others, 2015). Without using any ground control, DEMs were found to be accurate to within ± 30 cm at 95% confidence, with the vertical offset typically being the largest. A single high-quality ground control point could reduce this accuracy to the precision level of the DEM. The precision (repeatability) was about ± 8 cm at 95% confidence. In the coastal study presented here, we found that without any ground control the vertical accuracy could be several meters rather than several decimeters although horizontal accuracy was better than 30 cm. The reason for the large difference in vertical accuracy is likely due to the waves creating a motion parallax that gets interpreted as an elevation parallax. That is, the moving wave crests are used as matched points between photos by the software, which assumes the waves are not moving. One approach to reducing the

080

20/3/2015

6

error would be to crop the water out of each image before processing, but this is labor intensive task and unreasonable for long stretches of coast. Where ground control is available, as in our study, the DEMs can be shifted to those GCPs, again reducing accuracy to the precision level, as in prior work (Nolan and others, 2015) and as seen in Table 1, on the order of 10 cm. Because the GCPs are not used in processing of the DEMs, and the precision of the DEMs is so high regardless of DEM extents, there is theoretically no minimum spacing required for GCPs as a single one will suffice although a higher density will always improve QA/QC. If GCPs are not available in the study area, prior DEMs can be used for a similar purpose. If these also are unavailable, either cropping the ocean from the images before processing or processing some contiguous ocean-free location inland and bridging that control to the coast should result in the ± 30 cm directlygeoreferenced accuracy, though this remains to be tested. Regardless of GCPs or their spacing, the near-constant elevation of the instantaneous water line can be used both as a survey-bridge of near-constant elevation and as validation for the precision of the DEM, as any variations from this constant elevation could be attributable to spatially coherent or incoherent noise. The analyses presented here report changes relative to the co-registered data sets (after the image-image/DEM-DEM adjustment) regardless of their absolute (“real world”) coordinate systems and thus, in order to evaluate uncertainty in vertical or volumetric change we derive an empirical uncertainty based on measured volume changes over stable areas where no change would be expected between survey periods. On Barter Island, as for much of Alaska, the identification of stable features is difficult due to limited fixed infrastructure such as paved roads and sidewalks (the existing dirt and gravel roads and airstrip are continually graded), the seasonal growth of tundra vegetation, and/or regional changes in ground elevations associated with warming and permafrost layer thaw and subsidence. For this analysis we chose four separate areas on the gravel pads near the radar station. Within these areas we calculated an area-averaged volume change of 2 cm (range of 1-5 cm). When applied to our 27,000 m2 bluff change study area, an empirical uncertainty on our volume change calculation is ±540 m3, which corresponds to a 2% error by volume. Meteorological Data Hourly water levels and meteorological data, including wind speed and direction and barometric pressure measured at Prudhoe Bay (the only continuously recording tide gauge on the north coast of Alaska, ~180km west of Barter Island) and similar meteorological data measured at Barter Island airport were acquired for the time period July 1-Sep 30, 2014 (fig. 2). The diurnal tidal range along the

080

20/3/2015

7

north coast of Alaska as measured at Prudhoe Bay is about 0.21 m (NOAA, 2015), however, water levels can become elevated or depressed up to several meters due to winds and low-pressure systems; westerly winds tend to elevate water levels, whereas easterly wind events tend to lower water levels (Reimnitz and Maurer, 1979; Sultan and others, 2011; Erikson and others, in press).

Fig. 2 Meteorological and water-level data from Barter Island and Prudhoe Bay during the study period. Note the three periods of elevated water levels (red asterisk) between the two surveys (red lines). The high-water events generally correspond to periods of westerly winds and low barometric pressure (gray vertical bars).

Results and Discussion The high-resolution color orthoimages and DEMs developed for this study show a phenomenal level of detail of the landscape with exceptional detail of both large and small scale coastal features; wrack lines, high-water lines, berms, and sediment texture are all beautifully imaged (fig. 3). Comparison of the aerial imagery and derived DEMs show considerable bluff retreat and morphological change occurred along the Barter Island coast during the summer of 2014 (figs. 4 and 5). Mean change in the position of the top and base of the coastal bluff and the shoreline for the 2.7 km section of coast analyzed are -0.9 m, -2.9 m, and -2.4 m, respectively (Table 2). There was an associated net mean volume loss of

080

20/3/2015

8

approximately 28,000±540 m3 between the top and the base of the bluffs within the analysis extent. 75% of the loss (21,000 m3) occurred within zone 2, a 1.2 km long stretch of coast (55% of the analysis area) in the middle of the study area west of the snow fence (fig. 4). Perspective views of the bluffs show that the pattern of change was dominantly characterized by landward retreat of the top of the bluffs and removal of the debris fan at the base of the bluffs (fig. 5).

Fig.3. Examples of the detailed aerial orthophotos collected in September, 2014. A) this beach and gully located on the eastern edge of Barter Island’s bluffed coast shows a well-defined wrack/debris line on the upper beach and gravel material that was deposited on the lower beach during the early September storm; B) located west of the bluffs, this low-lying barrier spit is migrating landward over the inundated tundra and washover features from summer-2014 storms are well-defined. Old tundra/marsh deposits are visible in the middle part of the gravel beach face, along with the recent storm’s wet/dry line and multiple beach ridges deposited as the September water levels fell. The short dashed black lines shows the previous location of the adjacent logs in the July. Table 2. Mean linear change in across-shore feature location, in meters Feature

Minimum

Maximum

Mean

n

Top of bluff

0

4.3

0.9

255

Base of bluff

1

7.6

2.9

238

Shoreline

0

16.6

2.4

276

Although definitive information is lacking regarding the specific timing and drivers of the observed change, the water level and meteorological data recorded at Prudhoe Bay and Kaktovik indicate there were at least 3 high-water events

080

20/3/2015

9

associated with westerly winds and/or the passage of low-pressure systems during the study period (fig. 2). Elevated water levels along the Barter Island coast commonly results in flooding and overwash of low-lying areas and removal of sediment at the base of the permafrost bluffs triggering large-scale bluff failures. The authors were present at Barter Island toward the end of the third and largest storm which occurred just prior to the September data acquisition. Hourly averaged water levels measured at Prudhoe Bay during the storm were elevated nearly 90 cm above mean-sea level, which is over 4 times the diurnal tidal range. Although no ocean-water levels were measured at Barter Island during the storm, wind speeds measured at Kaktovik were on average 20 cm/s higher than at Prudhoe Bay suggesting water levels at Barter Island were at least as high or higher and fully inundated beaches and direct wave attack at the base of bluffs were observed during the storm (fig. 6). As water levels fell, most if not all of the debris at the base of bluffs was removed and the development of thermo-erosional niches at the base of ice-rich bluffs at some locations was observed. Whether all change can be attributed to that one storm or if similar conditions occurred during the prior two (or more) events is unknown but it’s clear that at least some of the removal of base material occurred in late-September. It’s likely that thermal (and mechanical) niche formation occurred at the peak water levels when the beaches were completely inundated and waves were able to directly impact the bluffs. Conclusions Aerial photography and coincident DEM development using the structure-frommotion (SfM) technique provides a highly detailed data set that is valuable for evaluating fine-scale change in coastal environments. Benefits of SfM include: 1) The methodology is relatively low cost (for example, as compared to traditional airborne lidar surveys) and provides high-quality, detailed data. 2) The combination of high-resolution elevation data with optical imagery is optimal for evaluating fine scale changes and a visual record of surface features. 3) Accurate measurement of bluff retreat and volume change. Relatively fine changes in bluff edge position (< 1m) were accurately measured along the 2.7 km stretch of bluffed coast and a cumulative volume 27,000 m3 of material was eroded from the bluff, primarily through the removal of debris at the base of the bluffs, with only about 2% error by volume. 4) The level of detail, especially in the September is outstanding. Wrack lines, high-water lines, berms, and sediment texture are all beautifully imaged.

080

20/3/2015

10

Additional considerations: 1) Integration with ground control points and groundtruthing is required to reduce absolute error and to integrate with existing geographic coordinate systems especially in coastal environments where movement of offshore waves can result in greater vertical errors in the DEM compared to settings with motionless ground features. 2) Because of the high spatial resolution of this type of data care must be taken to ensure that accurate geographic coordinate systems, with specific geodetic reference frames, are used in order to evaluate data accuracy, uncertainties, and valid change detection results.

Fig.4. Aerial photographs collected in A) July and B) September, 2014. Note in the July image the presence of ice offshore and ice and snow around snow fences, in gullies, and on the beach surface. C) DEM difference map of the two surfaces overlain on the September orthophoto; hot colors are erosion and cool colors are accretion. Note the substantial changes (retreat) to the bluffs in the center of the image, as well as snow melt around the snow fences and in the gullies. The inset box outlines the approximate view extent shown in perspective views in Figure 4 and the location of the cross shore profile a-a’.

080

20/3/2015

11

Fig. 5. Perspective views looking west for a section of the Barter Island coast showing an example of coastal bluff retreat observed at Barter Island during the summer of 2014; a) July imagery draped on the DEM; b) September imagery draped on the DEM; c) the change measured between the two periods; d) plan view of the area around the profile a-a’ (north is down); and e) cross section profile a-a’ for July 1 (red), and Sept. 8 (green) showing over 3 meters of retreat of the top of the bluff and considerable removal of the debris material at the base.

080

20/3/2015

12

Fig. 6. Photograph of the Barter Island bluffs during the early September 2014 storm, showing an inundated beach and relatively high waves that are actively eroding the base of the coastal bluff. Photo credit: Tom Lorenson, USGS.

Acknowledgements This study was funded by the U.S. Geological Survey, Coastal and Marine Geology Program as part of the National Assessment of Coastal Change and Coastal National Elevation Dataset Development projects. Thanks to Nicole Kinsman, Alaska DGGS, for thoughtful discussions and helpful review. Any use of trade, product, or firm names is for descriptive purposes only and does not imply endorsement by the U.S. Government. References Dolan, R.; Hayden, B., and Heywood, J., (1978), “A new photogrammetric method for determining shoreline erosion,” Coastal Engineering, 2, 21-39 Erikson, L.H., Gibbs, A.E., Richmond, B.M., Storlazzi, C.D., Jones, B.M., and Ohman, K.A., (in press), “Changing storm conditions in response to projected 21st century climate change scenarios and the potential impact on an Arctic barrier island–lagoon system—A pilot study for Arey Island and Lagoon, eastern Arctic Alaska,” U.S. Geological Survey Open File Report.

080

20/3/2015

13

Gibbs, A.E., Erikson, L.H., Jones, B., and Richmond, B.M. (2010). “Characterizing morphology and erosional trends of permafrost bluffs, Barter Island, Alaska,” 2010 AGU Fall Meeting Abstract EP23A-0772, San Francisco, Calif., 13-17 Dec. Gibbs, A.E., and Richmond, B.M. (2015). “National assessment of shoreline change: historical shoreline changes along the north coast of Alaska – U.S. Canadian Border to Icy Cape,” U.S. Geological Survey Open-File Report, 119 p. 71 figs. Hapke, C.J., (2005), “Estimation of regional material yield from coastal landslides based on historical digital terrain modelling,” Earth Surface Processes and Landforms, 30, 679-697. Jorgenson, M.T., and Shur, Y. (2008). “Glaciation of the Coastal Plain of Northern Alaska,” 2008 AGU Fall Meeting Abstract, San Francisco, Calif., 13-17 Dec. McBride, R.A., (1989), “Accurate computer mapping of coastal change: Bayou Lafourche shoreline, Louisiana, USA,” In: MAGOON, O.T.(ed.), Coastal Zone '89, (ASCE), pp. 707-719 Moore, L.J. (2000), “Shoreline Mapping Techniques.” Journal of Coastal Research, 16(1), 111-124. Nolan, M., Larsen, C.F., and Sturm, M. (2015). “Mapping snow-depth from manned-aircraft on landscape scales at centimeter resolution using Structurefrom-Motion photogrammetry,” The Cryoshpere Discuss. 9, 333-381. National Oceanic and Atmospheric Administration (2015), “Tides and currents— Prudhoe Bay, AK—Station ID 9497645”, National Oceanic and Atmospheric Administration Web site, accessed January 28, 2015, at http://tidesandcurrents.noaa.gov/stationhome.html?id=9497645. Reimnitz, E., and Maurer, D.K. (1979), “Effects of storm surges on the Beaufort Sea coast, northern Alaska,” Arctic, v. 32, no. 4, p. 329–344. State of Alaska, (2015), “Department of Commerce, Community, and Economic Development, Community and Region Affairs website, accessed January 28, 2015:http://commerce.state.ak.us/cra/DCRAExternal/community/Details/d3b 6b64d-90c9-4cec-8fe9-4d3b910a7b01” Sultan, N.J., Braun, K.W., and Thieman, D.S. (2011), “North Slope trends in sea level, storm frequency, duration and intensity,” Ice Tech Conference, Prudhoe Bay, Paper No. ICETECH10-155-R0.

080

20/3/2015

14

Thieler, E.R., Himmelstoss, E.A., Zichichi, J.L., and Ergul, A., (2009), “Digital Shoreline Analysis System (DSAS) version 4.0—An ArcGIS extension for calculating shoreline change,” U.S. Geological Survey Open-File Report 2008-1278. [Also available at http://woodshole.er.usgs.gov/projectpages/dsas/version4/.] U.S. Geological Survey, (2015), “AK_NORTHSLOPE lidar data set”, available at USGS EarthExplorer Web site, accessed January 28, 2015, at http://earthexplorer.usgs.gov.