Reduced sea-ice protection period increases storm exposure in Kivalina, Alaska

47 48 On Arctic coasts, erosion is limited by the presence of nearshore sea ice, which creates a 49 protective barrier from storms. In Kivalina, an Alaskan Inupiaq Inuit community, decreasing 50 seasonal sea-ice extent and a lengthening of the open-water season may be resulting in fall 51 storms that 1) generate higher, longer, and more destructive waves, and 2) cause damage later in 52 the year, resulting in increased flooding and erosion. We assess trends in the duration of 53 nearshore sea ice and their relationship with storm occurrence over the period 1979-2015 in 54 Kivalina. Analysis of passive microwave sea ice concentration data indicates that the open-water 55 season has increased by 5.6 ± 1.2 days/decade over the last 37 years, with moderate evidence 56 that it is extending further into the fall than into the spring. This is correlated with an increased 57 reporting frequency of high-damage storms; 80% of reported storms since 1970 occurred in the 58 last 15 years. Each high-damage storm event occurred during the open-water season for that 59 year. Our findings support Kivalina villagers’ assertions that climate change increases storm 60 exposure and associated damages from flooding and erosion. 61 63

The goal of this study is to characterize how changing sea-ice temporal patterns have 144 manifested in nearshore conditions along the western Alaskan coast. We assess trends in the 145 duration of nearshore sea ice and their relationship with storm occurrence over the period 1979-146 2015 in Kivalina. We hypothesize that climate change is reducing the extent and duration of 147 Arctic sea ice, and that shorefast sea ice -solid ice which forms along the coast and remains 148 attached to the shore (WMO 2014) -is forming later in the year than in previous decades, leaving 149 the village without protection from fall sea storm flooding and erosional wave action. about the Singuak Entrance, as evidenced by erosion of bars and shoals (USACE 2007). Beach 160 berms have accreted due to ice and storms, and are higher than the tundra further inland. Berms 161 consist of washed sand and gravel with material added by longshore drift and sediment 162 transported by rivers (Moore 1966) and provide the coastal tundra with some protection during 163 storms (USACE 2007 We selected six ocean grid cells adjacent to the Kivalina coastline, as defined by the 182 NSIDC land mask, from the data archive, covering a 125 km stretch of coast ( Figure 2). Due to 183 limitations from sensor resolution and the algorithm used, some land-to-ocean contamination is 184 D r a f t inherent, which may factor in the accuracy of the final analysis . To process the data, we  185  extracted the fractional sea-ice concentration (SIC) for each cell for every day data were taken  186  over a given year, over the period 1979-2015 for which the satellite record is complete. For each  187  cell, we determined first and last days of open water in a given year, defined as the first and last  188  days for which SIC was less than 15%; <15% SIC is commonly used as a marker of an ice-free  189 area  Table  216 1, R 2 =0.087, p<.001). As an illustration, there were 173 days of continuous open-water on 217 average across the six grid cells in 1987, whereas there were 184 in 2015.

219
Storm trends 220 The results of the literature review can be found in Table 2 The observed nearshore sea ice trends for Kivalina agree with a number of other studies, 227 which report declining Arctic sea ice to varying degrees. In general, the trends we observe are 228 consistent with, but not as pronounced as, other analyses. Cavalieri and Parkinson's (2012) pan-229 Arctic study using the NSIDC passive microwave data showed that all months exhibited negative 230 Chukchi, and western Beaufort region, the day of spring sea ice retreat was -1.5±0.2 days earlier 242 per year, while the day of fall advance was 1.3±0.2 days later. Overall, remote sensing data 243 indicate that sea-ice extent is declining across the Arctic, including along Alaskan coasts. 244 Varying results across the range of available evidence may reflect differences in 245 methodology when processing the sea ice data. More importantly, changes in sea ice extent 246 exhibit large spatial variability due to the complexity of Arctic atmosphere and ocean circulation 247 systems (Jeffries et. al 2013). For example, in an analysis of breakup and freeze-up dates for the 248 Chukchi Sea as a whole, Serreze et al. (2016) showed that variations about the trend line are 249 highly correlated with variability in the Bering Strait heat inflow, indicating that this inflow 250 plays a major role in controlling the dates of spring sea ice retreat and autumn ice advance. When 251 seasonal ice retreat occurs early, low-albedo open water areas are exposed earlier, which gain 252 energy from the sun; with more heat in the upper ocean, autumn ice growth is delayed. The 253 scatter about the trend lines in Figure 4, then, may to some extent be attributed to the highly 254 localized spatial extent but broad temporal extent of our study. 255 256 The literature review (Table 2) suggests a correlation between increased open-water 257 season and incidence of severe storms. As each high-damage storm event occurred during the 258 open-water season for that year, three interpretations are possible: (1) storminess has increased; 259 (2) the conditions in which damage occurs has changed; or (3) reporting frequency has increased. 260 For the first possibility, we cannot confidently infer changes in storm distribution, 261 because the data in Kivalina are limited to anecdotal accounts. Within the circumpolar sector, 262 Atkinson (2005) found that the Chukchi Sea experienced the highest-strength storms, with storm 263 power (defined as wind speed^2 * duration) reaching a minimum in July and maximum in 264 October; however, observed storminess was characterized by large inter-annual variability, and 265 no long-term trends were detectable in wind records from 1950-2000. Thus, we lack strong 266 evidence that storminess has changed significantly. This shifts emphasis to the second 267 possibility: Kivalina's physical vulnerability to impacts has increased due to an extended open-268 water season. 269 The third interpretation implies a change in risk perception, and increased urgency to 270 leave. This may also be the case, as the correlation between reduced sea ice and more frequent 271 severe storms reported here confirms local anecdotal observations. According to Colleen Swan, a 272 tribal administrator, "In the past the sea would freeze without fail (beginning in 273 October years. 337 • The OWS now extends two weeks later into fall than it did four decades ago. 338 • The longer open-water season has resulted in increased exposure to erosion and flooding, 339 with an increased frequency of high-damage storms. 340 • Erosion has been a concern in Kivalina since its founding in the early 20 th century. D r a f t Tables  Table 1. Slopes of the linear regressions for the first and last days of open water, and for the  length of the continuous open-water season, over the measurement period 1979-2015. There are  two options given below for calculating the open water season: (1) computing, for each grid cell, the first day when sea-ice concentration (SIC) went below 15%; or (2) computing the first day within the year when the average SIC across all six grid cells went below 15%. In the figures and text we depict the first option as more data were preserved when considering each grid cell individually, with R 2 =.005 (p=.31) for the change in the first day, R 2 =.185 (p<.001) for the change in the last day, and R 2 =.087 (p<.001) for the change in the overall length of the openwater season.

Input
Change in first day (days/decade) Change in last day (days/decade)

Change in continuous open-water season (days/decade)
SIC of each grid cell on each day (Fig 3) -0.9 4.7 5.6 Average SIC on each day -1.5 5.5 7.0 D r a f t The storm resulted in flooding of the lower portions of town, which was recorded at the port site as 5.05' above mean lower low water, likely from wave setup of 1' and run-up of 6.5' (Glenn Gray and Associates 2010b Federal government leaders had arrived in Kivalina to celebrate the completion of a multimillion-dollar, 1800' seawall composed of one-meter square metal baskets (gabions) filled with sand and rock. However, before the beginning of the celebration, a storm with winds exceeding 40 mph damaged 160' of the wall and forced the officials to cancel the celebration (Bronen and Chapin 2013). The storm eroded 50' inland, exposing permafrost in some areas.

12, 2007
A severe storm with waves up to 8' forced 221 residents to evacuate the village in search of safety (Gorokhovich and Leiserowizì 2012). Village leaders told the US Government Accountability Office that the evacuation was so dangerous that they would never again attempt an evacuation (Bronen and Chapin 2003).
Following this storm, the US Army Corps of Engineers approved the construction of a large rock revetment project, rising 13' above sea level, with a design life of 15-20 years. It was built in 2009-10.

Nov 8, 2011
A hurricane-strength storm occurred, causing 135 people to evacuate to the village school. However, the rock revetment succeeded in keeping out flood waters, so no flooding occurred, and no water damage was sustained.   727°N, -164.533°W). Kivalina is located on a narrow 10-13 km-long barrier island in the Chukchi Sea, 130 km north of the Arctic Circle; the nearest major city is Kotzebue .   Fig 2. Grid cells used in analysis of sea ice off the coast of Kivalina. NSIDC sea ice concentration data on October 26, 1978, rendered in Quantum GIS version 2.14.2-Essen. The figure directly represents the NSIDC data product, which has been packed into one-byte integers representing sea ice concentration values. A black pixel indicates zero sea ice, grey represents some fractional sea ice concentration, and a white pixel represents the superimposed land and coastline mask. A numerical grid system was used to obtain coordinates; grid cells are outlined and labelled by coordinate in inset. These cells cover a 125-km stretch of coast.   Table 1 shows the results of the linear regression.
(b) Changes in the first and last days of open water. The first and last days (lower and upper lines, respectively) were defined as the first and last days for which that grid cell's SIC was below 15%. The results for all of the grid cells are plotted, with each marker shape representing a different grid cell, over the measurement period 1979-2015. Table 1 shows the results of the linear regression. D r a f t Figures  Fig 1. D r a f t