Reconstruction of long-term hourly sea level data for Prince Edward Island, Canada

Abstract

Regional sea level rise has been regarded a key factor in understanding of climate change impact to coastal communities. As a vulnerable island to sea level rise and storm surges, the province of Prince Edward Island (PEI) in Canada lacks sufficient long-term island-wide historic record of sea level data. This has become a major challenge for further studies on coastal environments and climate change adaptation. To overcome this limitation, here we reconstruct a long-term hourly sea level dataset using the existing long-term records of limited permanent tide stations and short-term records of widely-distributed temporary stations. With comprehensive statistical analysis and modeling, the historical sea level records furthest between 1911 and 2023 are reconstructed with an hourly time step. This new dataset significantly extends the availability of long-term sea level data along with the shoreline of PEI, which can be used for further studies on coastal change assessment and coastal hazard adaptation in the context of climate change.

Background & Summary

Sea level rise (SLR) has been widely regarded as a threat to global coastal communities in the context of climate change¹. Previous research has revealed that the melting glaciers and increased buoyancy fluxes have been significantly lifting the sea level under a warming climate^2,3, which makes the coastal communities face the increasing risks of inundation and flooding. The research focusing on the 20^th century found that the averaged global SLR has changed from 1.7 to 3.3 mm/year, which has shown a significant accelerating trend from the middle to the late 20^th century⁴. At the regional-study scope, the Pearl River Estuary (by the north coast of South China Sea) reported a 3.9 mm/year SLR from 1993 to 2020⁵; the SLR in the North Sea has risen by 1.0 mm/year and has exceeded 2.7 mm/year after the year of 1993⁶; Southeastern Australia has shown an 0.2–0.3 m SLR in the last 200 years with an rising rate of 4.0 mm/year from 1900 to 1949⁷. In this context, there have been more frequent coastal hazards related to SLR globally⁸, such as coastal flooding events, long-term coastal land inundation, and coastal erosion. National wide observations across the USA have proved the growing frequency of coastal flooding events since the 1950s⁹; worldwide coastal communities, especially large cities, are facing a series of further impacts of urban inundation, ecosystem degradation, heritage destruction, and even displacement of population¹⁰; different regional studies have indicated that the coastal erosion will also be accelerated by the rising sea levels^11,12. Towards the various potential consequences associated with SLR, it is highly significant to have the record of the long-term and spatially-explicit regional SLR and to understand its changing pattern, for a resilient coastal flooding control and environmental management for coastal communities^5,13.

Prince Edward Island (PEI) is located by the south coast of Gulf of St. Lawrence, on the Atlantic coast of Canada (Fig. 1). The province of PEI is the smallest, lowest-lying and most densely-populated, and the only island provincial administrative region in Canada. In the recent decades, PEI has been increasingly reported having a rising vulnerability towards climate change impacts¹⁴, and has been regarded as a typical research site for climate change and adaptation studies, including the topics of coastal flooding^15,16, coastal erosion^17,18, community resilience^19,20, agricultural management^21,22, ecology and landcover^23,24, tourism management^25,26, and public health²⁷. These recent and ongoing regional research works in PEI are mostly to solve the challenges for the environmental security and sustainable development of the local coastal communities. The relevant challenges are tightly associated with the unique geographic conditions of PEI. Geologically, PEI has a low elevation range of 0–142 m with sedimentary geological bodies of red sandstone or mudstone, and red soil layer. Along the shoreline of PEI, a series of coastal morphologic areas have been formed by the ocean activities, such as estuaries, bays, beaches, bluffs, sandbanks, and barrier islands, which are basically in an unconsolidated texture¹⁷. PEI has a cool and humid oceanic climate (−3~−11C in winter and 20–34 C in summer) with over 1100 mm annual precipitation (about 80% rainfall and 20% snowfall)²⁸. Frequent storm hazards has occurred here in the recent 5 years, including 2019 Storm Dorian, 2021 Hurricane Ida, and 2022 Hurricane Fiona, which have all caused severe floods and erosions in the coastal environments and communities in PEI¹⁹. In this context, the only pillar industries of the PEI society, including agriculture, fisheries and tourism, have shown their sensitivities to the climate change impacts. Regional SLR around PEI, in the past and future, has been revealed and projected to significantly influence the coastal-change processes and further leading to adverse effects on the coastal land resources, local populations and socioeconomic activities^18,29. Especially the hazards of coastal erosion, storm surge, and flooding have been listed as the major climate change risks in PEI¹⁴, to which SLR may have an important contribution by raising the frequency of and the damage caused by these potential hazards, according to the global research experience^30–32.

Fig. 1.

PEI and surrounding sea areas on the terrain map (retrieved from ESRI via ArcGIS Pro), and the DFO tide stations (marked with ID) utilized in this study.

With the necessary need of understanding the regional SLR, the long-term and site-specific sea level records become an important reference³³. The previous studies in the Thames Estuary, UK¹³, coastal sites in Spain^34,35, estuaries in Northeastern USA^36,37, and Marseille, France³⁸ have shown researchers’ efforts in providing the datasets of regional sea level record. Along the shoreline of Canada, the Fisheries and Oceans Canada (DFO) has been providing the freely-available sea level records via Canadian Tides and Water Levels Data Archive (https://www.isdm-gdsi.gc.ca/isdm-gdsi/twl-mne/index-eng.htm)³⁹, in which the earliest record started from the year of 1848 and over 60 stations have had the historic record for over 50 years. While for the research site of PEI, there is only one permanent DFO tide station (i.e., Charlottetown tide station, station ID: 1700; in Fig. 1)⁴⁰, which started recording in 1911 and now is still in operation. Meanwhile, there used to be several temporary stations along the shoreline of PEI (Fig. 1), which run shortly in different historic periods and have short-term records of their running, but none of them are still in operation nowadays. In the recent years, the University of PEI has established the PEI Storm Surge Early Warning System (PSSEWS, https://pssews.peiclimate.ca/)⁴¹, which has set up multiple island-wide tide stations and has kept the sea level records after the year of 2020. However, for the other parts of the island, there still lacks long-term observation records of SLR.

Facing the need of keeping the long-term spatially-explicit sea level records and understanding the SLR around PEI across the recent two centuries, and the limitation of lacking island-wide sea level observations, we expect to explore the possibility of using the existing records of both permanent stations close to PEI (not limited to be in PEI) and the temporary stations in PEI to reconstruct the sea level variation in a correlation approach. This simulation will markedly improve the availability of the island-wide SLR records into decades scale throughout the 20^th century and the early 21^st century with reliable reconstructed sea-level data. The outcomes of this sea level reconstruction are expected to help the researchers focusing on PEI’s nearshore hydrodynamics, coastal erosion, maritime management, coastal landscape evolution, and regional historical geography. An open-source dataset of PEI regional SLR may also further provide available resources for decision-makers and the public to understand the coastal-change processes around the island and the impact of climate change.

Methods

Based on the existing records of the permanent and temporary stations around PEI, trying to establish the statistical correlation between them and then to reconstruct the sea levels can be considered as a feasible method to address the limitation of short record at the temporary stations. The core principle behind this method is that the same-time tide records of a temporary station have a high similarity with a nearby and geographically-connected permanent station, as the ebb and flow of the tides at these station combinations share the same processes in a certain spatial area^13,42–44. In this case, if the records have any temporal overlaps between a temporary and a permanent station, a linear regression can be generated based on that similarity, and the simulated long-term sea levels can be reconstructed following the linear correlation for each temporary station. According to the records of different DFO stations around PEI, their greatest common divisor of their sampling intervals in different periods is one hour³⁹. Hence, the hourly data, which we use in the reconstruction, also provides a relatively high temporal-resolution for getting more accurate correlations based on the record overlaps from monthly to interannual scales. Our reconstruction following three steps (Fig. 2), which are (1) examining the spectrum of each temporary station and its reference station in a randomly-selected 14-day window in their overlapped period (also shown in Fig. 4) to determine the temporal offset for regression, (2) establishing the correlation between each temporary station and its reference station using linear regression method and checking the R² value, and (3) reconstructing the long-term sea level data for each temporary station using the equation from step (2). It should be noted that due to the tidal dynamics, the best temporal matching may not always be the synchronous one in each combination. In other words, although the temporary station and its reference station has a highly-similar trend of sea level variation, their variation can be at the same time or a few (usually less that 3) hours different, which will be identified in the step (1).

Fig. 2.

Intermediate results from the three steps of sea level reconstruction (demonstrated at Wood Island (ID: 1680) station): (a) spectrum of the temporary station and its reference station; (b) linear regression of the matched water level records, after identifying the record of Wood Island best matching that 1-hour-earlier of Charlottetown; and (c) the comparison of the reconstructed sea level data and the correlated observation data.

Fig. 4.

Spectrum check for all the selected temporary stations.

Through our initial examination of the existing records of the permanent stations in or close to PEI, there are in total four stations appropriate for our study, which are Charlottetown (ID: 1700) in PEI, Shediac Bay (ID: 1805) in the province of New Brunswick, Cap-aux-Meules (ID: 1970) in the province of Quebec, and Lower Escuminac (ID: 2000) in the province of New Brunswick (Fig. 1, Table 1). All the four permanent stations are still in operation and have long-term records across decades since the last century, among which the Charlottetown station has the longest record since the year of 1911 and the continuous record since 1938 (Fig. 3). As for the temporary stations, there used to be in total 11 stations in PEI, however, 4 of them had poor records due to probable operational failure or could not satisfactorily match a permanent station with valid records based on the correlation, so that be discarded for the island-wide SLR reconstruction, which are Miminegash (ID: 1855), Tignish (ID: 1875), Alberton (ID: 1885), and Naufrage (ID: 1945). The other 7 temporary stations are selected for the reconstruction (shown in Figs. 1, 3 and Table 1).

Table 1.

Information of the applied stations.

Station	Station ID	Latitude (N)	Longitude (W)	Permanent/Temporary	Reference Station	Correlation	R2	Time difference
Charlottetown	1700	46.23	63.12	Permanent	√ (CH)	N/A
Shediac Bay	1805	46.23	64.55		√ (SB)
Cap-aux-Meules	1970	47.38	61.86		√ (CAM)
Lower Escuminac	2000	47.08	64.88		√ (LE)
Souris	1650	46.35	62.25	Temporary	CAM	y = 1.3217x + 0.2049	0.858	Synchronous
Wood Island	1680	45.95	62.75		CH	y = 0.7565x + 0.1381	0.936	1 hour earlier than CH
Port Borden	1725	46.25	63.70		CH	y = 0.7921x + 0.2276	0.892	Synchronous
West Point	1845	46.63	64.40		SB	y = 0.8692x −0.2026	0.882	Synchronous
Goodwood River	1896	46.62	63.92		LE	y = 0.8534x + 0.3942	0.897	2 hours later than LE
Rustico	1915	46.47	63.28		LE	y = 0.5682x + 0.2261	0.798	2 hours later than LE
North Lake Harbor	1955	46.47	62.07		CAM	y = 0.8810x + 0.7838	0.814	Synchronous

Fig. 3.

Duration of datasets at all stations around PEI.

Through checking the sea level records of all the stations, there are some unavoidable data points losses found in the records of the selected stations. Therefore, a manual examination and correction has also been done for all the overlapped data points of each station combination, to ensure that the overlapped records of the two stations best match each other. The final reference stations are determined by the highest coefficient of determination (R²) values among the comparisons in addition to the spectrum check (Fig. 4). There are in total 3 stations having temporal shifts applied in the reconstruction progress, which are the stations of Wood Island (ID: 1680), Goodwood River (ID: 1896) and Rustico (ID: 1915), they cannot match any permanent stations synchronously. The possible reasons of the time offsets can be special geomorphologies, such as estuaries, bays, river channels and straits, that the tidal current may travel for some time to reach the farther direction^13,43. We have concluded the best asynchronous matching of the lagged correction for these three stations, which are Wood Island best matching 1-hour-earlier Charlottetown, and Goodwood River and Rustico best matching 2-hour-later Lower Escuminac.

With the linear correlation obtained for each selected temporary station and its reference station (shown in Table 1), the simulated sea levels at the locations of all the selected stations can be reconstructed using the correlation to transform the sea level records of the reference stations. Note, there are also some unavoidable data losses in the records of the selected permanent stations, which could result from equipment failure, extreme weather, and equipment maintenance. As the availability of the record of the simulated reconstruction for temporary stations fully depends on their reference stations, if the reference stations have any lost data points, they will be reflected in the simulated data for the associated temporary stations.

Data Records

This newly generated sea level reconstruction for the nearshore area around PEI, Canada is freely available to the public via the Figshare database (10.6084/m9.figshare.27989570)⁴⁵. The dataset consists of 13 files, including one introduction (TXT) file named “ReadMe.txt”, one Excel spreadsheets (XLSX) file named “PEI_sea_level_reconstruction.xlsx” containing sea level records for all the selected 11 stations, and 11 comma separate variable (CSV) files named in the form of “Station ID_Name.csv” for respectively for each selected station. The temporal resolution for all the data is one hour, and the time zone is the local time (UTC-4). The datum used in the dataset is the chart datum (CD), that represents the reference plane to which depths on a published chart, all tide height predictions, and most water level measurements are referred⁴⁶. The unit used for water level is meter (m) and each value is accurate to two decimal places (i.e., cm). For different permanent stations, Charlottetown station has the earliest record in Jan. 1911 and continuous records since 1938. Shediac Bay station has the earliest record from Nov. 1971 and has a one-month (July) break in 1981 and an interannual break from Mar. 1992 to Aug. 2003. Cap-aux-Meules station has the earlies record in Mar. 1964 while has several on-and-off breaks between Dec. 1969 and June 2007, which may be due to its high difficulty for maintenance on an offshore island. Lower Escuminac has the continuous records since Jan. 1973. Accordingly, the associated temporary stations have the same duration of available reconstructed data as their own reference stations. Through this reconstruction work, the long-term sea level data in PEI has been expanded from the only available station of Charlottetown to in total 8 stations across the island (Fig. 5). The newly-generated dataset is expect to play a supportive role in relevant studies in the coastal environments of PEI, especially for those may need a historic record as a reference.

Fig. 5.

Distribution of the reconstructed sea level data in PEI.

Technical Validation

The objective of establishing this reconstructed sea level dataset is to maximum address the lack of observation records. Hence, it is not feasible to directly compare the data points with any observed reference values. In this case, it is more important to check if this simulation dataset is reasonable or reliable enough for providing the SLR changing pattern. Accordingly, two key questions should be noted for the technical validation: 1) if the expected matched data points of each station combination are used for the linear regression due to the existing losses of data points; 2) if the linear correlation between each selected temporary station and its reference station is reliable enough.

For the matching examination, we have manually and visually gone through each overlapped sea-level match throughout all the 7 station combinations, which contains in total 46,152 matches. The asynchronous adjustments for Wood Island, Goodwood River and Rustico were made after the temporal-matching corrections. For the reliabilities of the linear correlations, we have examined the R² values of them. The results showed that all of them exceeded 0.79, and 5 of the 7 exceeded 0.85 (shown in Table 1). Although there is no fixed requirement for R² value, we would like to justify that our results have been operated the spectrum examination (Fig. 4) for all the station combinations to ensure the reconstructions are based on the correlated sea level variation trends, and some research experiences in oceanography have even used significantly lower values (e.g., 0.30–0.60) as their threshold for approval and have regarded the R² values over 0.70 as highly-correlated^47,48. Therefore, we regard our results having a strong reliability to well reflect the sea level variations around PEI as a simulated dataset. While in the meantime, the future users of this dataset should also notice the potential bias between the simulated sea-levels and the objectively real ones that is generated from random errors that cannot be addressed through our reconstruction, which may be from the differences in local-scale short-term hydrometeorological processes like extreme rainfall or wind-wave setup, or the operational issue of individual tidal gauge.

Author contributions

T.P. collected the original data, analyzed and modeled the statistical correlation, generated the new dataset, wrote the first draft of the manuscript. X.W. originated the concept of this study, supervised the data collection, analyses and reconstruction, and contributed to the manuscript revision. M.Q.M. and S.B. contributed to data examination and correction, and manuscript revision.

Code availability

Our newly-generated dataset can be found on the open-source database, Figshare (10.6084/m9.figshare.27989570). There is no specific code produced to support this effort. All the analyses described in this paper can be accomplished by commonly-used statistical tools like Microsoft Excel or IBM SPSS Statistics. The original data for this reconstruction work can be downloaded via Canadian Tides and Water Levels Data Archive (https://www.isdm-gdsi.gc.ca/isdm-gdsi/twl-mne/index-eng.htm).

Competing interests

The authors declare no competing interests.