Recent increase in catastrophic tropical cyclone flooding in coastal North Carolina, USA: Long-term observations suggest a regime shift

Coastal North Carolina, USA, has experienced three extreme tropical cyclone-driven flood events since 1999, causing catastrophic human impacts from flooding and leading to major alterations of water quality, biogeochemistry, and ecological conditions. The apparent increased frequency and magnitudes of such events led us to question whether this is just coincidence or whether we are witnessing a regime shift in tropical cyclone flooding and associated ecosystem impacts. Examination of continuous rainfall records for coastal NC since 1898 reveals a period of unprecedentedly high precipitation since the late-1990’s, and a trend toward increasingly high precipitation associated with tropical cyclones over the last 120 years. We posit that this trend, which is consistent with observations elsewhere, represents a recent regime shift with major ramifications for hydrology, carbon and nutrient cycling, water and habitat quality and resourcefulness of Mid-Atlantic and possibly other USA coastal regions.


Monitoring and sampling.
Biogeochemical and water quality data were obtained from the long-term ModMon monitoring programs in the NRE and western PS 10 . ModMon is a collaborative University -State of North Carolina (NC-Dept. of Environmental Quality-DEQ), and UNC-Chapel Hill Institute of Marine Sciences (IMS) program (http://paerllab.web.unc.edu/projects/modmon/), initiated in 1994. Sampling consisted of twice monthly visits to 11 mid-river stations along the estuarine portion of NRE (Fig. 2), including vertical profiles with collections of near-surface and near-bottom water for physical-chemical-biological parameters. Monthly samples were also collected at nine stations in the western PS as part of the ModMon program starting in 2000 (Fig. 2). Profiles of temperature, salinity, and dissolved oxygen were made at 0.5 m depth intervals using YSI 6600 multi-parameter water quality sondes (Yellow Springs Inc, Yellow Springs, Ohio). Sondes were calibrated prior to each sampling trip according to the YSI User's Manual.  . Dissolved organic nitrogen (DON) was computed by difference as TDN -DIN. Details on sample preparation and processing are in 17 . Filtrates were analyzed for dissolved N forms and SRP with a Lachat/Zellweger Analytics QuickChem 8000 flow injection autoanalyzer using standard protocols (Lachat method numbers 31-107-04-1-C, 31-107-06-1-B, and 31-115-01-3-C, respectively) 17 . Particulate organic carbon (POC) and nitrogen (PON) were measured on seston collected on pre-combusted GF/F filters, analyzed by high-temperature combustion using a Costech ECS 4010 analyzer 18 . DIC and DOC were measured on a Shimadzu Total Organic Carbon Analyzer (TOC-5000A) 19 .
Phytoplankton biomass. Chlorophyll a (Chl-a) was measured for near-surface and near-bottom samples by filtering 50 mL of NRE water onto GF/F filters. Filters were frozen at −20 °C and subsequently extracted using a tissue grinder in 90% acetone 17 . Chl-a of extracts was measured using the non-acidification method of Welschmeyer 20 , on a Turner Designs Trilogy fluorometer calibrated with pure Chl-a standards (Turner Designs, Sunnyvale, CA).
Freshwater discharge, and material loading. Daily average Neuse River discharge was measured by the United States Geological Survey (USGS) at Fort Barnwell (USGS 02091814), and divided by 0.69 to account for ungaged downstream inputs 17 . Daily Neuse River loads of carbon and nutrient forms were estimated using Weighted Regressions on Time Discharge and Season (WRTDS) 21,22 , based on daily average discharge and concentrations measured by ModMon (or NC DEQ for total N and total P) at the head of the estuary (Fig. 2). Half-window widths of the tricube weight function for seasonality, time, and discharge were set to default values of 6 months, 7 years, and 2 natural log units, respectively 21 .
A long term record of precipitation events in the Neuse River basin at Kinston, NC was assembled from National Oceanic and Atmospheric Administration Cooperative Observer Network sites 314684 (1 September 1899 to 15 June 2017) and site 314689 (15 June 2017 to 5 December 2018). Data from 1 May 1919 to 29 November 1923 were not available. Precipitation events were defined as daily precipitation greater than 4.85 cm, the 99 th percentile of daily precipitation. Consecutive days of rainfall greater than 4.85 cm were considered the same event and assigned to the day the event began. Precipitation events were ascribed to tropical cyclones when a precipitation event was coincident with the passage of a tropical cyclone within 240 km (150 miles) of Kinston, NC as determined by 6 h storm advisories recorded in the National Hurricane Center's HURDAT2 database. Quantile regressions were constructed for the 90 th and 50 th quantiles of cyclone related precipitation against time. 95% confidence intervals on slopes for the quantile regressions were determined by 1000 rounds of bootstrapping. A long term record (1 May 1930 to 8 December 2018) of high river flow events for the Neuse River was assembled from the USGS gage at Kinston, NC (USGS gage 02089500). High flow events were those during which the daily average flow was greater than 390 m 3 s −1 , the 99 th percentile of daily average flow. Consecutive days of flow greater than 390 m 3 s −1 were considered the same event and summed from the day the event began to determine the www.nature.com/scientificreports www.nature.com/scientificreports/ total event discharge. Discharge events were ascribed to tropical cyclones when a discharge event followed within a seven-day period the passage of a tropical cyclone within 240 km (150 miles) of Kinston, NC. Discharge was regulated by Falls Lake dam from 1982 to the present.

Results and Discussion
Extreme rainfall associated with recent tropical cyclones. With less than a 2% chance of three such events occurring in a twenty-year period 23 , either North Carolina has been very unlucky, or the historical record used to define the storm statistics is no longer representative of the present climatic regime. This analysis suggests that the occurrence of three extreme floods resulting from high rainfall tropical cyclone events in the past 20 years is a consequence of the increased moisture carrying capacity of tropical cyclones due to the warming climate 4,6,24-26 .
While we do not offer a full attribution analysis, which may be conducted in a variety of ways including numerical modeling that replicates the events 27,28 , our observations are consistent with observations elsewhere and with predicted changes in a warming climate 2-6 . Moreover, rather than attributing a particular event to global warming, we should consider whether a warming climate made these events more likely, which our records suggest is the case for coastal NC. For example, increased precipitation in other US coastal areas subject to tropical cyclones (e.g., coastal Texas from Hurricane Harvey in 2017) [29][30][31] and increased hurricane activity since 1970 30 have been attributed to global warming. Factors potentially driving the increased precipitation include; (1) greater heat content of ocean waters, which not only fuels storm intensity but also increases precipitation 29 , (2) a decrease in tropical cyclone forward movement 32 providing more opportunity for heavy precipitation over a particular area, (e.g. Harvey and Florence), (3) an observed poleward migration of tropical cyclones 33 , perhaps making coastal NC more vulnerable than in the past, and (4) an increase in tropical cyclone intensity in the satellite era 34 .
Fortunately, North Carolina has a well-kept continuous record of tropical cyclone landfalls and associated rainfall since 1898, which we investigated in order to further test the hypothesis that we have recently entered a regime shift of increased extreme rainfall and associated flooding. Three periods of elevated cyclone activity were noted; the first in the early 1900's (~1910), then during the 1950's and most recently since the mid-1990's (Fig. 3A). However, six of the seven highest precipitation events, four of the six due to tropical cyclones, have occurred in the past 20 years. Both the median and 90% quantile of precipitation from cyclone-related extreme precipitation events have increased significantly over the past century (Fig. 3B), and a more rapid increase in the 90% quantile reflects the recent occurrence of those six very high precipitation events. In addition, these events have been accompanied by record freshwater discharge to the NRE (Fig. 3C). The rank of the total event discharge www.nature.com/scientificreports www.nature.com/scientificreports/ in Fig. 3 does not necessarily correspond to its rank in peak flow. For example, total event discharge following Hurricanes Floyd (1999) and Florence (2018) were significantly higher than following Hurricane Matthew (2018) which had the highest instantaneous peak flows and flood stage. Part of this result was due to continued heavy rainfall following hurricanes Floyd (including Hurricane Irene) and Florence, and following Florence was also due to retention and slow release of flood waters through Falls Lake dam upstream from the NRE.
Biogeochemical and water quality impacts. The most recent extreme precipitation events each have delivered up to 100 cm of rainfall to coastal watersheds; often accounting for 30-45% of the average annual rainfall 8 . The floodwaters resulting from Hurricane Florence (~14 September, 2018) completely "freshened" the entire NRE, leaving only an oxygen poor salt wedge downstream at its mouth, near the entrance to Pamlico Sound (Fig. 4). The floodwaters contained extremely high loads of organic matter 35 , dominated by dissolved organic carbon (DOC) and dissolved organic nitrogen (DON) as well as nutrients, specifically the growth-limiting nutrient nitrogen 36 , shown here as NO x , (Fig. 5). In addition to depressing salinity throughout the NRE, the floodwater nutrient load fueled phytoplankton production and subsequent algal blooms (as chlorophyll a), which were promoted after the flow rates (and hence flushing) decreased, enabling phytoplankton biomass to build up in the estuary (Fig. 6). Phytoplankton blooms and associated hypoxia often continued for days to weeks after a storm had passed. Analysis of C:N content of the dissolved and particulate organic matter entering the NRE (Fig. 7) indicated that a vast proportion of the DOM have quite high C:N ratios (>20), consistent with terrigenously-derived sources that were evident from satellite images of the NC coast (Fig. 1). Degradation of this terrigenous DOM by a combination of sunlight and bacteria may have kept the Pamlico Sound as a net CO 2 source to the atmosphere in the weeks following Hurricane Matthew 35 . In contrast, POM had C:N ratios <10 (Fig. 7), indicating authochtonous sources, most likely phytoplankton production, in this eutrophic estuary. Sustained primary production has the potential to modulate CO 2 dynamics by creating a CO 2 sink 18,35 .
Thus, evidence is accumulating that we may also be seeing changes to the "system state" of coastal waters in terms of their ability to capture or release CO 2 37,38 . Such changes caused by an increased frequency of extreme storm events are ostensibly reorganizing coastal carbon cycles 38 . For example, flood waters reaching the inner shelf of the Gulf of Mexico have resulted in extensive degradation of terrestrial organic matter and the return of that carbon as CO 2 to the atmosphere 38 . Further, the biogeochemistry of coastal waters is inextricably linked to their water quality 9,15 . Floodwater-associated nutrients have been shown to promote harmful algal blooms (HABs) in these systems 39 . Floodwaters contain contaminants and runoff from urban and agricultural land use 13 , and due to the high organic load, flood waters are often hypoxic when they enter an estuary, which was www.nature.com/scientificreports www.nature.com/scientificreports/ evident in the %DO values < 40% in the upper NRE following each storm (Fig. 4). Additionally, high freshwater inflows reduce vertical mixing and "trap" denser salt water causing extensive hypoxia in bottom waters extending throughout the estuary (Fig. 4). These hypoxic events can last weeks to months and provide the ingredients for massive finfish and shellfish kills, as well as an abrupt increase in fish disease 36,40 Losses suffered by coastal communities from these events can be catastrophic. All of the aforementioned biogeochemical effects have severe economic and societal implications for fisheries, tourism, and real-estate, and have raised concerns about coastal resiliency and sustainability. In North Carolina alone, Hurricane Floyd in 1999 caused fisheries losses of US$6 million and overall economic (tourism, property and business damage and losses, agriculture and silviculture) amounted to US$2 billion 41 .
While the hydrologic, nutrient and carbon inputs attributable to Florence (Sept.-Nov. 2018) are yet to be fully tallied, the rainfall associated with this event was roughly equivalent to Matthew, in 2016. Like Floyd and Matthew, Florence's floodwaters led to "freshening" and expanding hypoxic zones in the APS system (Fig. 4), as  www.nature.com/scientificreports www.nature.com/scientificreports/ well as massive pulses of carbon overflowing from the APS into coastal waters, as viewed from space ( Fig. 1), with effects that can linger for months after a storm 35 .

Conclusions
Considering these extreme precipitation events and their hydrologic and biogeochemical consequences in totality, it is clear that they are unparalleled in the past 120+ years of recorded tropical cyclones in coastal North Carolina (Fig. 3). The potential exists for receiving waters globally to undergo unprecedented perturbations to nutrient and carbon cycling, fisheries habitat and sustainability due to increasing frequency of extreme precipitation events; all of which are still to be determined. With roughly 40% of the world's population within 100 km of the coast, development inland, as well as along the coastline, will exacerbate the perturbations caused by this type of regime shift 42 . We stress that stakeholders, state and federal governments need to better prepare for the acute as well as cumulative water quality, fisheries resource and overall socio-economic effects of this recently-documented rise in catastrophic flooding associated with elevated tropical storm activity.