**4. Discussion**

The combination of a historical aerial photograph review and shoreline change analysis provides a semi-quantitative summary of the impacts of storms on the Napatree Barrier. This combination of datasets allows for a review of historical storm events and the resulting impacts on the barrier. The results highlight the importance of storm timing and frequency relative to the timescale of foredune recovery. Based on storm surge elevation and regional coastal impacts, the most substantial storm to impact Napatree and the shoreline change analysis during the study period, was the 1938 hurricane. Washover fans are clearly visible in both oblique (Figures 5 and 9) and vertical aerial photographs following the 1938 hurricane. The post-storm photographs and 1939 vertical aerial images show complete removal of the foredune (and houses) along the barrier. While the historical photographs show clear impacts to the barrier, the net retreat of the shoreline over this period is less clear. Individual transects between 1883 and 1939 show changes ranging from essentially no change (+/−<3 m) to >20 m of retreat near the western end of the barrier (Figure 6). The average change along the barrier between 1883 and 1939 was −9.4 m. This falls within the positional uncertainty of the shoreline pair (+/−13.4 m; Table 4), so it cannot be said conclusively that the hurricane of 1938 produced net migration of the shoreline. The spring 1939 photographs were taken 236 days after the storm, so it is likely the active beach

had largely recovered, and the high-water line position would not be considered storm impacted. The 1938 hurricane was likely the first time the barrier had been overwashed in at least several decades, so the dunes were likely well developed prior to the storm, and a vegetated foredune is visible in the 1934 and 1936 images (Figures 5 and 9).

Aerial photographs were not collected between 1939 and the 1944 hurricane, so the condition of the barrier and levels of dune recovery and revegetation cannot be assessed. Dune recovery when a barrier is completely overwashed can take up to a decade [65], so complete recovery following the 1938 hurricane was unlikely prior to the 1944 hurricane. Vertical aerial photographs from 1945 show that Napatree was overwashed presumably during the 1944 hurricane. The 1945 imagery was not used as part of the shoreline change analysis, as the photos were 'washed out,' and delineating a high-water line was problematic. Between 1939 and 1948, the shoreline retreat exceeded the positional uncertainty of the shorelines, with an average change of −24.1 m and a maximum change of −30.7 m (Table 4). The measured shoreline change, coupled with the washover fans visible in the 1945 imagery suggests that the 1944 hurricane resulted in a net migration of the barrier. A similar response was seen in 1951 imagery and the 1948 to 1951 shoreline pair. Fresh washover fans are visible in 1951 imagery, and the shoreline change between 1948 and 1951 averaged −13.4 m (the same value as the positional uncertainty); however, some transects exceeded −22 m of change (Table 4). Water levels at the Newport gauge did not exceed the storm threshold between 1948 and 1951. An extra-tropical storm in late November 1950 produced a 0.85 m storm surge; however, the storm peaked after high tide, so the maximum water level recorded was 0.5 m MHHW at the Newport tide gauge. Water levels were higher to the west of Napatree, peaking at 1.4 m MHHW at New London and 0.9 m MHHW at Montauk, suggesting that impacts of this storm were more substantial in western Rhode Island and eastern Connecticut. Elsewhere, the November 1950 extra-tropical storm produced washover fans and caused substantial damage along the New Jersey coastline [43]. Following the hurricane of 1944 coupled with an extra-tropical storm in 1947, foredune recovery was limited and the barrier remained low enough to be overwashed during this storm.

Aerial images were not collected between 1951 and 1962, so the impacts observed over that period are the cumulative results of Hurricanes Carol and Edna (1954), Hurricane Donna (1960), an extra-tropical storm in 1953 and the Ash Wednesday Storm (1962), among other smaller extra-tropical storms. As a result, the impacts of individual storms cannot be parsed out; however, the net result was migration of the barrier that exceeded the positional uncertainty. A shoreline was not derived from the 1962 shoreline, as it was considered too soon after the Ash Wednesday storm (39 days); however, between 1951 and 1963 the shoreline retreated an average of −21.9 m, with a maximum change of −38.9 m (Table 4). The timing of these storms following the 1950 extra-tropical storm continued to hinder foredune recovery, and each storm event that overwashed the barrier essentially reset the clock on dune recovery. The shoreline continued to retreat between 1963 and 1975 (average change −10.1 m; maximum change −15.7 m); however, the aerial photography shows that vegetation was largely reestablished at this point and migration via overwash had largely ceased, and it appears the barrier was narrowing. Barrier width increased between 1883 and 1963 and decreased between 1963 and 2016. The combination of shoreline change data and observations from the aerial imagery suggests that the increase in width was likely driven by the overwash of the barrier and deposition of washover fans on the back-barrier. Increased width via overwash and washover fan deposition was also noted in field surveys after the 1938 hurricane [26].

While the 1938 hurricane produced appeared to produce little net change in shoreline position compared to 1883, storms' impacts on the barrier were the impetus for the migration of the barrier that occurred between 1939 and 1975. The removal of the foredune increased the susceptibility of the barrier to overtopping in future storms, which allowed subsequent storms that likely would not have overwashed the barrier to overtop the partially recovered foredune, and deposit washover fans on the back barrier, 'rolling

over' the barrier. The combination of field and LiDAR measurements provides insight into the elevation recovery from a moderate event (Hurricane Sandy). The majority of the Napatree barrier just exceeds the pre-Sandy volume and elevation 5.5 years after the storm, suggesting recovery from this event where the foredune was not completely removed took ~5 years [65]. Most of the vegetation was reestablished on the back barrier within 1 year following Sandy (as seen in 2014 aerial images and 2013 field surveys), and this pattern was likely repeated in smaller past storms, where washover was limited to a few 10s of meters across the barrier and dune vegetation can reestablish quickly.

No storm exceeded the threshold at the Newport tide gauge between November 1963 and December 1974. This 11-year period apparently allowed the dunes to recover enough elevation and volume to prevent wide-spread overwash of the barrier in subsequent moderate storm events. Based on the rates of recovery of the dune measured since Sandy [65], the foredune could have recovered ~1 m of elevation and ~30 m3 m−<sup>1</sup> of volume during that period. This is supported by work on other barriers, where recovery periods of <5 years to >10 years have been reported following storms that completely overwash the barrier [64,66,67]. It remains unclear if management practices (i.e., sand fences) were utilized in the study area in the past, and if so, how they impacted dune formation and recovery between 1963 and 1975. Sand fences have clearly been used along the eastern end of the barrier over the last few decades, with multiple levels to build the foredune up to 8 m (MLLW) just west of the groins. The overall increase in dune height from west to east along the barrier was likely driven by a combination of some sand fencing and sediment availability and the prevailing wind direction; the dominant wind directions are southwest (spring/summer) and northwest (fall/winter), both of which transport sand towards the eastern end of the spit. This likely explains the partial recovery and revegetation of the eastern 500 m of the barrier prior to 1975, which limited overwash in these areas. There is no record of beach replenishment at this site; the exception to this is filling of the inlet breach at the eastern end of the barrier following the 1938 hurricane.

Subsequent storms, including the Blizzard of 1978, which caused extensive erosion elsewhere in New England, did not appear to cause widespread overwash, and no localized overwash was apparent in 1981 aerial images (See supplemental materials). A similar response was observed for Hurricane Gloria (1985), which overwashed other areas of the RISS [53], and the combined impact of Hurricane Bob and the October 1991 extra-tropical storm, which had limited overwash of the dunes and localized washover fan penetration of 10 s of meters. Hurricane Sandy (October 2012) had a similar storm surge elevation as the 1944 hurricane, yet it in little change in position of the barrier. Overwash was limited to the western portion of the barrier, washover fans only extended across the barrier in a few locations and the eastern portion of the barrier remained largely in the collisional regime of Sallenger [68]. The 1944 hurricane had a greater impact because the barrier (specifically the foredune) had not recovered from the 1938 hurricane; the same storm event, if the foredune had fully recovered, would likely not have produced the same impacts on the barrier. Houser et al. [64,66] showed that a more erosive storm being followed by smaller storms can produce a larger cumulative impact and result in substantial shoreline retreat. The 1938 to 1975 period falls into the 'best-case' scenario of Houser et al. [64], where storms were clustered, followed by periods of recovery. Under these conditions, transgression is rapid, and dunes are small and discontinuous during the periods of storminess; however, the subsequent period of quiescence allows for the dunes to reestablish, and eventually return to pre-storm height [64]. Fenster and Dolan [7,69] reported a reversal in shoreline change trends, from a seaward migration to more landward migration between 1930 and 1970. This period of increased shoreline change was attributed to storm frequency—notably, an increase in extra-tropical storm frequency. The trend reversed (switched from erosional to accretional) around 1967–1968 [69], which falls between available aerial photographs for Napatree (1963 and 1972/1975). Donnelly et al. [43] reported a similar response to what occurred with the storms discussed here—overwash and washover fan deposition on back barrier salt marshes of Brigantine Island in New Jersey (250 km west of Napatree) in 1944, 1950 and 1962. This suggests that the paraglacial isolated and welded Napatree barrier behaved similarly to the barrier island chains along the Mid-Atlantic coast of the U.S.

Shoreline change rates calculated between 1939 and 1975 were higher than previously published for the Napatree Barrier (Table 1). The annualized rates calculated by Boothroyd et al. [23] between 1939 and 2014 and the USGS long term rate between 1883 and 2004 [25] are less than half the annualized rate observed at Napatree between 1939 and 1975 (Table 1). The USGS assessment [25] also included a short-term shoreline change rate for the period between 1975–2000, with reported shoreline change values reflecting progradation of the shoreline. The average (end point) rate was +0.4 m yr−<sup>1</sup> for that period (Table 1). Taken together, this suggests that the long-term shoreline change rates underestimate actual rates of change, which are an order of magnitude higher following a large storm event. While long-term shoreline change rates may produce a better mean representation of the longterm trend [70], the observed shoreline migration between 1939 and 1975 are indicative of the change that can occur during a shorter interval with several impactful storms. This does not suggest that the storms are outliers; rather, if the shoreline migration occurs during a cluster of storms, the annualized rate of change is reduced by spreading the change out over a longer time with limited storm activity.

Coastal construction setbacks in Rhode Island are based on the shoreline change rates between 1939 and 2014. Specifically, the setback was calculated as 30× the long-term erosion rate. These rates do not include additional components to account for storm impacts [71]. The results of this work show that on this shoreline these rates are likely underestimated if a barrier experiences a series of significant storm events and can exceed the long-term trend here by a factor of three or more. Zhang et al. [6] argued that including storm-influenced shorelines in long-term shoreline change calculations leads to an overestimation of rates of shoreline change and could have consequences on coastal development based on setbacks. However, where storms drive the observed rates of change and dominate the shoreline change signal, the inverse would potentially be of more consequence, as setbacks calculated using low rates of change fail to calculate the actual risk to coastal properties over time. While the Napatree barrier is currently undeveloped, a hypothetical house constructed using the average long-term (1883–2014) shoreline change rate (−0.6 m yr<sup>−</sup>1) would face substantially more risk if a similar storm sequence that was observed beginning in 1938 were to occur. Separating the shoreline response to storms from the long-term shoreline change record and understanding the risk to coastal development remain challenges for coastal managers. Numerous studies have outlined the likely response of barriers to sea-level rise [12–15,17,47]. These impacts include an increased rate of landward migration, more frequent overwash with the rising sea level and increased storminess, increased frequency of breaching and inlet formation/widening. The period between 1938 and 1975 provides at least a partial analog for the potential responses of other barrier spits within the glaciated northeast to a period of increased storminess. Headland separated barriers occur elsewhere within New England in both wave dominated and mixed energy regimes, and along Nova Scotia and Northern Germany [28]. Understanding the responses of these types of barrier systems to storms and a rising sea level fills an important gap in knowledge, as much of the existing literature focuses on barrier islands, particularly along the mid-Atlantic coast of the U.S.

A conclusive cause and effect summary linking all the observed shoreline changes to individual storms is not possible given the positional uncertainty and short-term changes observed. The lack of randomization, replication and direct measurements following storms also limit direct cause and effect attribution of the shoreline change to individual storms. However, the combination of shoreline change analysis and historical aerial photography here suggests that much of the change observed was the result of washover fan deposition during storms, which led to migration of the barrier. Extrapolating the shoreline trend between 1883 and 1939 (dashed grey line in Figure 7), which is admittedly limited to two data points, or the period between 1975 and 2014, suggests that the subsequent shoreline positions have not returned to the long-term trend. The interpretation here is

that the storms produced net migration of the shoreline. If the change in the shoreline was driven by some other long-term factor (i.e., sea-level rise or changes in sediment supply), the shoreline should return to the long-term trend following storms [6]. Higher rates of sealevel rise over the last 40 years [39] do not coincide with any measurable shoreline retreat, and even correspond with a period of progradation (within the positional uncertainty) between 1975 and 2014. The responses of barriers under various sea-level rise scenarios have been the focus of recent modelling efforts, and the dynamics of the foredune have been shown in models to impact barrier retreat—namely, when the dunes are low, barriers are more susceptible to overwash, leading to migration of the barrier [21]. This leads to episodic retreat of the barrier, followed by periods of relative stability in barrier positions when the foredune recovers sufficiently to limit overwash. These results, both shown here at Napatree, reinforce the relationship between dune recovery and storm frequency reported in recent modelling studies [21].

Discussion of storm impacts here does not negate the impact of sea-level rise on barriers; however, the cumulative impact of storms between 1938 and 1975 likely far outweighs any response due to sea-level rise. The transgression observed over the span of a few decades could be indicative of future behavior of headland separated barriers if storm frequency increases in the future. Relative sea-level rise increases the vulnerability of the dunes to erosion, and this would exacerbate the impacts of a similar stormy period in the future. This combination of storms and sea-level rise could cause the barrier to cross a geomorphic threshold and lead to faster rates of transgression. Under moderate to high rates of sea-level rise and given the relatively consistent (low) back-barrier slope, the Napatree barrier could experience 'width drowning' in which the rapid overwash and migration of the barrier would outpace sediment transport from the shoreface, leading to barrier narrowing and possible barrier loss over the next few centuries [15]. Napatree, like the rest of the RISS, is sediment-starved, with little modern sediment on the shoreface [32]. Incision into the underlying glacial deposits is possible; however, this process has not yet been performed for this shoreline. Due to lacking an abundant sediment supply, Napatree is susceptible to increased storm frequency/overwash. An additional sediment added to the barrier is from either erosion of the small bluff at the east end and/or transport from the shoreface and back barrier during transgression. Transport of the sediment down the Pawcatuck River into Little Narragansett Bay is likely not a significant contributor of sediment to the Napatree Barrier [72].

What the morphology of the Napatree barrier will look like if continued transgression causes the spit to detach from the western headland remains unclear. A possible response may be for the barrier to rotate clockwise to a more northeast to southwest orientation, similarly to the migration of the Sandy Point barrier (Figure 1). This would somewhat follow the model of Orford et al. [73], where barriers between glacial till headlands are breached, and switch from swash-aligned ones to ones where the barrier is oblique to the headland and rapid migration is driven by longshore sediment transport. Future research should focus on coring and geophysical studies to examine the properties of the shoreface, and observations of physical processes coupled with models would be helpful to quantify sediment transport pathways on the shoreface. Sediment transport on the northern side of the barrier, including exchange between the barrier and deposits related to the former position of Sandy Point, and sediment lost from the system around the west end of Napatree Point, remains unknown. While Napatree lacks significant back barrier marsh and is not part of a restricted coastal lagoon where loss of marsh due to accelerated sea-level rise may significantly alter tidal characteristics [17], inlet formation and widening could affect sediment distribution, including the transport of sediment from the barrier and adjacent shoreface to a flood-tidal delta [11]. Transport of sediment into the tidal delta represents loss to the subaerial barrier system, as the sediment remains there until transgression of the barrier reaches the tidal delta [11]. This loss of sediment would be problematic for a starved system, particularly if an inlet remained open long enough to develop a flood-tidal delta. The overall response of the barrier here is illustrative of the challenges barriers

face in periods of increased storminess, and the importance of storm frequency relative to dune recovery. It represents an important case study when considering storms' impacts on isolated, mainland attached barriers in future climate models.

**Supplementary Materials:** The following are available online at https://www.mdpi.com/article/ 10.3390/geosciences11080330/s1. Figure S1: Monthly sea-level elevations at the Newport, RI, New London, CT and Montauk, NY tide gauges. Figure S2: 1934 Partial vertical aerial photograph of the Napatree Barrier. Figure S3: 1939 Vertical aerial photograph of the Napatree Barrier. Figure S4: 1945 Vertical aerial photograph of the Napatree Barrier. Figure S5: 1951 Vertical aerial photograph of the Napatree Barrier. Figure S6: April 1962 Vertical aerial photograph of the Napatree Barrier. Figure S7: September 1963 Vertical aerial photograph of the Napatree Barrier. Original 9 × 9 Photograph scanned by Boothroyd and Hehre. Figure S8: April 1972 Vertical aerial photograph of the Napatree Barrier. Figure S9: April 1975 Vertical aerial photograph of the Napatree Barrier. Original 9 × 9 Photograph scanned by Boothroyd and Hehre. Figure S10: 1976 Vertical aerial photograph of the Napatree Barrier. Figure S11: April 1981 Vertical aerial photograph of the Napatree Barrier. Figure S12: March 1985 Vertical aerial photograph of the Napatree Barrier. Original 9 × 9 Scanned by Boothroyd and Hehre, 2007. Figure S13: April 1988 Vertical aerial photograph of the Napatree Barrier. Figure S14: March 1992 Vertical aerial photograph of the Napatree Barrier. Figure S15: Spring 1997 Digital Orthophotograph of the Napatree Barrier. Figure S16: April 2004 Digital Orthophotograph of the Napatree Barrier. Figure S17: Spring 2008 Digital aerial photograph of the Napatree Barrier. Figure S18: June 2012 Digital Orthophotograph of the Napatree Barrier. Figure S19: April 2014 Digital Orthophotograph of the Napatree Barrier. Figure S20: April 2018 Digital Aerial Photograph of the Napatree Barrier. Table S1: Water levels exceeding the 0.7 m MHHW threshold at the Newport, RI and New London tide gauges. Table S2: Average distance from baseline and standard deviation of shoreline position used in this analysis. See Table 3 for the sources of the shorelines.

**Funding:** This research received no external funding. Publication costs were provided via a Connecticut State University-American Association of University Professors Faculty Research Grant. The Watch Hill Conservancy provided in-kind support for field surveys.

**Acknowledgments:** This manuscript benefited from numerous discussions over the years regarding shoreline change mapping and processes, including Mark Borrelli, Rachel Henderson, Janet Freedman, Robert Hollis and Scott Rasmussen. Mark Borelli, Peter August, Janet Freedman and Nathan Vinhateiro, along with two anonymous reviewers provided helpful reviews of the manuscript.

**Conflicts of Interest:** The authors declare no conflict of interest. The funders had no role in the design of the study; in the collection, analyses, or interpretation of data; in the writing of the manuscript or in the decision to publish the results.
