Analysis of a 60 km segment of the Alaskan Beaufort Sea coast using a time-series of aerial photography revealed that mean annual erosion rates increased from 6.8 m a−1 (1955 to 1979), to 8.7 m a−1 (1979 to 2002), to 13.6 m a−1 (2002 to 2007). We also observed that spatial patterns of erosion have become more uniform across shoreline types with different degrees of ice-richness. Further, during the remainder of the 2007 ice-free season 25 m of erosion occurred locally, in the absence of a westerly storm event. Concurrent arctic changes potentially responsible for this shift in the rate and pattern of land loss include declining sea ice extent, increasing summertime sea surface temperature, rising sea-level, and increases in storm power and corresponding wave action. Taken together, these factors may be leading to a new regime of ocean-land interactions that are repositioning and reshaping the Arctic coastline.
 Rates of coastal erosion in the Arctic are known to be among the highest in the world due in part to the ice-bonded nature of the coastal sediments [Reimnitz et al., 1988; Jorgenson and Brown, 2005]. Historically, mean annual erosion rates along the Beaufort Sea coast in Alaska are as high as 8 m a−1 for exposed ice-rich bluffs [Jorgenson and Brown, 2005] (Figure 1). The coastline here has a northern exposure with bluffs up to 6 masl consisting of ice-rich, near shore marine, glacio-fluvial, alluvial, and aeolian deposits that lack protective barrier islands [Reimnitz et al., 1988; Jorgenson and Brown, 2005]—an ideal setting to support very high erosion rates and an opportunity to study the direct effects of changing Arctic conditions along the ocean-land interface.
 The Beaufort Sea is traditionally ice-free for three to four months of the year, and it is during this short time period that all coastline erosion occurs. In general, erosion of the ice-rich bluffs located along this coastline involves the formation of a thermo-mechanic erosional niche, collapse of bluff materials once niche propagation exceeds bluff strength, deposition of the failed block at the bluff toe, and mass wasting of the block through thermal and mechanical degradation [Reimnitz et al., 1988]. The number of thermo-mechanical niche forming episodes per annum, and thus the erosion rate, is a function of ice-free season duration, the number and type of storms impacting the coastline, sea-level, and summertime sea surface temperature (SST). Localized spatial variation in niche formation is affected by bluff height, degree of ice-richness, and sediment composition, as well as near-shore bathymetry and the presence of barrier islands [Reimnitz et al., 1988; Jorgenson and Brown, 2005; Solomon, 2005]. Further, this landscape is characterized by large, elongated thermokarst lakes, with 70% of the land surface affected by thermokarst lake processes [Hinkel et al., 2005], creating much of the spatial variability in ice-richness along the coast.
 Understanding contemporary erosion rates is important because amplified Arctic climate change is leading to rapid and complex environmental responses in both terrestrial and marine ecosystems [Richter-Menge et al., 2008], including record reductions in Arctic sea ice extent [Stroeve et al., 2008], sea-level rise [Richter-Menge et al., 2008], warming of sea-surface [Steele et al., 2008] and permafrost [Brown and Romanovsky, 2008] temperatures, and increasing terrestrial permafrost degradation [Jorgenson et al., 2006]; all of which are potential drivers that may affect the rate and pattern of coastline erosion in the Arctic. Any increases in already rapid rates of coastal retreat will have ramifications on Arctic landscapes that provide important freshwater and terrestrial wildlife habitat [Flint et al., 2008], subsistence grounds for local communities, the loss of cultural sites that archive human settlement in the Arctic [Jones et al., 2008] represent loci for resource extraction infrastructure [Houseknecht and Bird, 2006], and potentially impact the global carbon budget by transferring organic carbon from terrestrial to marine storage zones [Hayes et al., 2007]. To quantify and better understand how the terrestrial-marine interface has responded to changing environmental conditions over the past 52 years, we delineated a time-series of coastlines from available aerial photography to measure erosion rates for a portion of the Beaufort Sea coast in Alaska.
 In this study, remote sensing and Geographic Information System (GIS) techniques were combined to delineate coastline positions from available historic and contemporary aerial photography, archived by the U.S. Geological Survey, Earth Resource Observation and Science (EROS) data center, from 15 August 1955 (1:55,000 scale, B&W), 19 July 1979 (1:63,360 scale, CIR), and 18 July 2002 (1:40,000 scale, CIR), and acquired recent imagery on 17 July 2007 (1:50,000 scale, true-color) for a 60 km segment of exposed, north-facing coastline along the Alaskan Beaufort Sea coast. The 2002 imagery were orthorectified and served as the base in order to geo-correct the other datasets at a spatial resolution of 2.5 m with mean RMS values of 2.6, 3.0, and 2.4 m, for image mosaics from 1955, 1979, and 2007, respectively. The coastline was defined as the ocean-land interface and delineated using a semi-automated classification technique described by Jones et al. . Erosion rates were determined for the three available time periods using the U.S. Geological Survey, Digital Shoreline Analysis System (DSAS) [Thieler et al., 2005] at 100 m increments along the 60 km segment of coastline for each time period. The dilution of accuracy (DOA) associated with the coastal retreat rates was estimated by (equation (1)):
where Eg is the positional accuracy of the 2002 orthophotos (determined to be ±5 m with differential global positioning system data collection), Ep1 and Ep2 represent the pixel resolution of the imagery from a particular year, RMS1 and RMS2 are the root mean square errors associated with the registration of an image mosaic from a particular year, and ΔT is the time interval for a given time period (modified from Hapke  and Lantuit and Pollard ). Thus, the annualized error associated with the erosion rate measurements is ±0.3, ±0.3, and ±1.1 m a−1 for the time periods 1955 to 1979, 1979 to 2002, and 2002 to 2007, respectively.
 To compare potential factors causing changes in erosion rates for these time periods, we compiled available datasets describing storm events, directional fetch relative to sea ice extent, and sea surface temperatures. Storm events for the late-open water season (August–October) were determined from the terrestrial observational record at Barrow, Alaska from 1955 to 2007. Storm events with the ability to perform geomorphological work were defined by wind speeds greater than 10 m s−1 for a minimum duration of 6 hr [Atkinson, 2005]. Effective storm power values (speed2*duration) were calculated by summing core storm events for wind directions toward the study coastline for each of our time periods. Ice-pack mediated fetch was measured for the same five wind directions from Lonely, Alaska using median September ice extent positions derived from microwave data from 1979 to 2007 to represent relative open-water fetch distances [Fetterer et al., 2002]. Summertime SST data for the western Beaufort Sea, calculated on a yearly basis for July–September, and defined as the mean temperature over the upper 10 m, were compiled by Steele et al. .
3. Results and Discussion
 Results reveal that mean annual erosion rates for the study coastline increased from 6.8 m a−1 (1955 to 1979), to 8.7 m a−1 (1979 to 2002), to 13.6 m a−1 (2002 to 2007) (Figures 2a–2c), indicating that erosion rates may be accelerating. In addition, 24% of the coastline exhibited rates greater than 10 m a−1 in the first time period, whereas this number increased to 43% and 65% in the 1979–2002 and 2002–2007 periods, respectively. Our erosion rate measurements for the earlier time periods agree quite well with rates compiled through previous investigations [Reimnitz et al., 1988; Jorgenson and Brown, 2005], indicating confidence in the imagery and methods used. We also better resolve recent observations that suggest that land area lost doubled relative to historic patterns beginning in the 1980s [Mars and Houseknecht, 2007]. Our more detailed spatial and temporal analysis reveals that it has occurred more recently. Further, observations of coastal retreat along the Beaufort Sea coast west of our study area near Barrow also indicate increased erosion since 2003 [Aguire et al., 2008].
 Not only have erosion rates accelerated recently, but during image analysis it appeared that a shift occurred in the magnitude of erosion rates along bluffs consisting of different degrees of ice-richness. Thus, we divided the study coastline into four distinct shoreline types based principally on thermokarst lake processes and corresponding vegetation and edaphic characteristics [Hinkel et al., 2003; Jorgenson and Shur, 2007] (Figure 2d). In the first two time periods, erosion rates were highest (12.5–13.5 m a−1) for the youngest land surface, recently drained lakes with basins lacking a vegetative mat and with relatively low ground-ice content (Type I). Erosion rates at older, ice-rich surfaces were substantially lower during the 1955–1979 time period (6.5–7.3 m a−1); this included lakes drained sufficiently long ago to allow revegetation and ground-ice enrichment of the basin (Type II), and land surfaces not affected by thermokarst lake activity with very high ground-ice content (Type III). However, during the last five years, measurements of coastal retreat reveal that the three different shoreline types are eroding at nearly identical rates, roughly 18 m a−1 (Figure 3).
 Differential erosion rates of these coastline types during previous periods indicate that shoreline characteristics play an important role in moderating coastal erosion. Recently tapped lake basins have traditionally exhibited the greatest erosion rates, which we attributed to low bluff heights with reduced mass for removal by transport, a relatively narrow beach, and the thawed condition of bank sediments to the waterline or below. Under such conditions, erosion can occur over a greater range of wave energy, where erosion via thermo-mechanical niche formation tends not to be the dominant process. In contrast, the greatest increase in erosion rates occurred for the ice-rich upland terrain, from 7.3 m a−1 (1955–1979) to 18.3 m a−1 (2002–2007), suggesting a fundamental shift in the processes driving and resisting erosion. It is therefore critical to consider the shifting balance of the forces necessary to cause such a dramatic shift.
 The prevailing wind direction along this portion of the Beaufort Sea coast is northeasterly, yet storm events traditionally have winds from westerly and northwesterly directions, while the most severe storms on record at Barrow during our study period occurred in 1963 and 2000 [Lynch et al., 2003]. However, wind events along the Arctic coast, those exceeding 10 m s−1 for a duration of at least 6 h are believed to have the ability to perform geomorphological work [Atkinson, 2005], though this relationship may vary considerably with direction, fetch, local sea-level, and SST. Comparison of storm power value means by study period show an increase of 59% from 1955–1978 to 1979–2001 and 35% from 1979–2001 to 2002–2006 with notable increases in easterly wind events between the first and second periods (Table 1), though these events are believed to be fairly ineffective in terms of erosion. The analysis of changes in fetch shows a doubling in open-water expanse in the northwest direction between the second and third time period (Table 1) due to declining sea ice extent, potentially creating more effective storm events.
Table 1. Sea Surface Temperature Anomaly, Storm Power Value, and Fetch Data From 1955 to 2007
Wind direction of storms directly affecting the study coastline.
SPv is storm power value as annual cumulative mean.
F is directional fetch (km) relative to median September ice pack position.
 A similarly important factor with respect to the erosive effectiveness of wind events and those periods not characterized by storms are the recent trends towards increased SST and sea-level rise. Summertime SST generally cooled in the western Beaufort Sea from 1955 to 1965, warming slightly at the end of our first time period, followed by a gradual increase throughout our second time period, with pronounced warming since 2000 and positive temperature anomalies of 2.5°C in 2007 [Steele et al., 2008] (Table 1). Sea-level rise on the order of 2.61±0.47 mm a−1 has also been documented along Arctic coastlines from 1954–2007, also with an abrupt increase between 2000–2007 [Richter-Menge et al., 2008]. Thus, these recent trends towards warming SST and sea-level rise may act to weaken the permafrost dominated coastline by preferential thawing of ice-rich coastal bluffs.
 During the remainder of the 2007 ice-free season we were also able to assess erosion for a small segment of coast along an ice-rich permafrost bluff (Figure 4). We found that 25 m of erosion occurred locally, in the absence of a westerly or northwesterly wind event, but with very high storm power values with easterly winds (Table 1). This indicates one of three possibilities: 1) that wind events from easterly directions are becoming more effective, 2) westerly winds of lesser magnitude, which we did not capture in our storm power value analysis, are becoming more effective due to increased open-water extent 3) or recent trends in Arctic Ocean surface warming and sea-level rise may be contributing.
 We hypothesize that it is the interaction of each of these factors that has led to the increased rate and uniformity of coastline erosion, contributing to more rapid thermo-mechanical niche development, fragmentation into blocks with more surface area, and denudation of the block material. Such shifts may potentially explain the 25 m of erosion observed here in 2007 and the disproportionate increase in erosion rates along ice-rich permafrost bluffs relative to ice-poor bluffs between 2002 and 2007. However, a longer time period and more detailed analyses are needed to assess the relative contribution of changes in each of the forces driving and resisting erosion. For instance, the recent patterns documented in our study may simply represent a short term episode of enhanced erosion. Alternatively, they may represent the future trajectory of coastline erosion in the Arctic. In either case, understanding the potential causes of such shifts in the rate and pattern of erosion is of interest for predicting arctic environmental response to climate change, particularly as large portions of the Arctic are being targeted for additional hydrocarbon development. Rapid erosion since 2002 resulted in the loss of a late-1970s petroleum test well (J.W. Dalton) and further erosion will soon result in the loss of the Drew Point test well (Figure S1 of the auxiliary material). More knowledge of the changes in the patterns and rates of coastline erosion will be needed to develop appropriate strategies and structures for onshore and offshore development.
 We thank D.V. Derksen, C.J. Markon, and R.A. Beck for their contributions to the research; K. Bollinger, H. Wilson, and R. MacDonald for field assistance and logistical support; and T. Ravens, C. Ely, D. Douglas, L. Gaydos, L. Hollands-Bartels, J. Brown, and one anonymous reviewer for their careful and critical reviews. Research funding was provided by the U.S. Geological Survey, Alaska Science Center.