Marine heatwaves (MHWs) have been the subject of increasing attention as their frequency and duration have increased over much of the ocean in connection with rising sea surface temperatures (SST) (Oliver et al., 2021). Much of this attention stems from the sensitivity of many marine ecosystems to temperature stress (Smale et al., 2019) and observations of ecosystem damage caused by MHWs (Arias-Ortiz et al., 2018; Cavanaugh et al., 2019; Aoki et al., 2021). This paper examines MHWs in the shallow coastal bays along a portion of the U.S. mid-Atlantic coast between Chesapeake Bay and Delaware Bay ( Figure 1 ). A 2015 MHW caused significant dieback in a large restored seagrass (Zostera marina) meadow in one of the bays (Berger et al., in review). Portions of the meadow closer to a tidal inlet experienced less dieback and recovered more quickly than portions further from the inlet (Aoki et al., 2021), likely because tidal exchange of cooler ocean water mitigated impacts of the MHW in regions of the bay proximal to the inlet.

A full analysis of MHWs requires long-term records of daily average water temperature to define climate normals (30-yr records are standard (WMO, 2017)) and episodic significant deviations from those normals. Such long-term instrumental records are relatively rare in shallow coastal systems. In this study, a 40-year record (1982-2021) of daily average water temperature in the study bays was constructed from several sources of measured water temperature. A similar instrumental record was developed for a nearby site in the coastal ocean. An attractive alternative to the lack of availability of long-term instrumental surface water temperature records in marine environments is provided by gridded and gap-free global SST datasets such as the NOAA Optimum Interpolation Sea Surface Temperature Version 2.1 (OISST-V2) 0.25°-resolution dataset of daily SST that spans the period September 1981-present (Reynolds et al., 2007; Huang et al., 2021) and the ESA Sea Surface Temperature Climate Change Initiative SST CCI and CS3 (SST-CCI/CS3) 0.05°-resolution datasets of daily SST, covering the same period of time (Merchant et al., 2019). Other high resolution datasets include OSTIA 0.05°-resolution daily SST (Good et al., 2020), which is available for 2007 – present, and CoralTemp 5-km daily nighttime SST (Liu et al., 2014), which extends from 1985 – present.

These gridded datasets were developed using satellite, buoy, and ship-based water temperature data together with a variety of calibrations and bias corrections to yield best agreement with observed SST (Banzon et al., 2016; Good et al., 2019; Good et al., 2020; Liu et al., 2014). Error estimates provided as part of the OISST-V2 set of data products show an increase in potential error in coastal regions of the ocean. Because many of the world’s most productive and threatened marine ecosystems are found in coastal environments (Millennium Ecosystem Assessment, 2005), the potential for errors in identification and characterization of marine heatwave events in coastal ecosystems associated with errors in gridded SST datasets in these regions is a cause for concern. Comparison with measured temperature records and higher resolution satellite-derived SST data during specific events can provide insight into the accuracy of gridded, gap-free data products at shallow coastal sites.

In this study, long-term (40-year) instrumental records and gridded SST datasets were used to investigate temperature variations and MHWs in a system of shallow coastal bays on the mid-Atlantic coast of the USA, and their relationship to MHWs in the adjacent coastal ocean. A MHW in June 2015, documented to have caused seagrass dieback in one of the coastal bays, provides a useful case study for comparison of datasets across sites and sources. Simulations of SST and MHWs at sites in the bay and coastal ocean from 1982 to 2100 illustrate the challenges temperature-sensitive coastal ecosystems will face in the future as water and air temperatures continue to increase.

Long-term, complete, gridded sea surface temperature (SST) datasets have transformed the study of marine heatwaves (MHWs), leading to scores of recent studies; see Oliver et al. (2021) for a recent review. Measured and gridded SST time series at the coastal ocean site were relatively well correlated ( Tables S3 , S4 ). Differences in means and trends of SST and MHW metrics derived from the measured (COC-Meas) and OISST-V2 (COC-OI) datasets for the coastal ocean site were not significant and both differed in consistent ways from SST and MHW metrics and trends calculated using the measured time series at the coastal bay site (CBW-Meas) ( Table 2 ; Figure 5 ).

It is important, however, to recognize the limitations of these datasets. In the context of the present study of marine heatwaves in shallow coastal bays, the resolution of the NOAA OISST-V2 dataset is too coarse to resolve bays that are less than a few hundred km² in surface area. The SST-CCI/CS3 and OSTIA datasets have a higher resolution, but the results shown in Figures 1 , S6 as well as Tables S1–S4 indicate that the SST-CCI and OSTIA datasets did not accurately capture high summer SST in the relatively small, shallow coastal bays characteristic of the study area and many other coastal-barrier systems. This is problematic because shallow coastal bays and lagoons are home to many important coastal ecosystems, some of which are known to be sensitive to heat stress (e.g., seagrasses (Berger et al., in review) and oysters (Donelan et al., 2021)). Few coastal bay sites have the long-term records of water temperature necessary to quantify MHW metrics and trends, making available gridded datasets an obvious alternative. The results presented here indicate use of these long-term high-resolution gridded datasets may significantly underestimate summer SST and lead to errors in identification of MHWs in relatively small, shallow coastal bays.

The 40-yr record of measured water temperatures constructed in this study for a site in the Virginia coastal bays illustrates the persistent, large differences in SST that can be present between shallow coastal bays and the adjacent coastal ocean. Water temperatures in the bays averaged several degrees warmer in summer and colder in winter than in the coastal ocean and temperature extremes were reached several weeks earlier ( Figure 2A ). The trend in SST was in the range of 0.2-0.3°C/decade at both sites ( Table 2 ), but the variance in daily average SST anomaly was twice as large at the coastal bay site ( Table 2 ; Figure 2B ). These differences reflect a greater responsiveness of water temperature in the coastal bays to changes in air temperature, as would be expected given their shallow depths. In addition, despite their proximity, air temperature at the coastal bay site, which is on the landward margin of the bays, was more variable than at the coastal ocean site, with an average summertime daily maximum-minimum air temperature difference of 6.5°C compared to a daily maximum-minimum difference of 1.5°C at the coastal ocean site ( Figure S3 ). The daily temperature metric most similar between summer air temperature and SST at the coastal ocean site was daily minimum while for the coastal bay site it was daily maximum ( Figure S3 ) suggesting that maximum daily air temperatures may be the primary control on water temperature in the shallow bays in summer whereas minimum daily air temperature is more important for SST in the coastal ocean.

Marine heatwaves are defined relative to local conditions. Higher temperatures in shallow coastal bays produce higher seasonal climatologies and MHW threshold conditions than those that apply to the coastal ocean site ( Figures 2A , S4 ). As a result, values of MHW exposure and intensity do not reflect the mean temperature difference between the sites. Annual MHW exposure was relatively similar at both sites from 1982-2021, although for 4 of the last 10 years, the coastal ocean site experienced dramatically higher annual exposure than in the previous 30 years ( Figure S8 ), mostly due to an increase in the frequency of events rather than their duration ( Figure 4 ). Annual intensity was consistently higher at the coastal bay site ( Table 2 ; Figure S8 ) because of the higher variance of the SST anomaly at that site. The high variance also affected the durations of exceedance events. Heatwave thresholds are typically defined such that, on average, 10% of days exceed the threshold temperature. Most exceedance events were not long enough to qualify as MHWs at either site ( Figure S7 ), but the fraction of events lasting 1-4 days was higher in the measured (75%) than the gridded (63%) coastal ocean record and higher in the bay (82%) than the coastal ocean, consistent with differences in variance among the records ( Table 2 ). By the same token, 5% of exceedance events in the ocean records lasted more than 20 days (measured and gridded), but just 1 of 417 exceedance events in the bay record was longer than 20 days ( Figure S7 ).

The maximum MHW threshold SST was 3°C higher at the bay site than the coastal ocean site ( Figure S4 ). The summer (June-August (JJA)) threshold difference was such that bay water temperature exceeded the ocean site MHW threshold for most of each summer: 78% of JJA, 71% of May – Sep. This was true even for relatively cool summers with no MHWs at the coastal bay or ocean sites. In contrast, ocean SST never exceeded the JJA bay MHW threshold during 1982-2021. This leads to the striking separation of peak summer MHW SST at the ocean and bay sites at a daily average temperature of 28.7-28.8°C ( Figure 4A ). This is close to the threshold temperature that has been associated with thermal stress in seagrass meadows (28.6°C) (Berger et al., in review). Because coastal ocean SST has been below this threshold even during MHWs, ocean water entering the bays through tidal inlets has provided some thermal relief to ecosystems proximal to the inlets (Berger et al., in review). For example, the June 2015 MHW caused significant dieback in the center of a seagrass meadow in one of the Virginia coastal bays (South Bay), while portions of the meadow closest to the inlet either experienced no dieback or recovered faster than portions further from the inlet (Aoki et al., 2021).

The temperature disparity between the bays and coastal ocean is also evident in the number of days each year when daily average SST ≥ 28°C ( Figure 5 ). Until the early 2000s, daily average SST never exceeded 28°C at the coastal ocean site; since then, there have been an average of less than 1.5 days/yr with SST ≥ 28°C. In contrast, the coastal bay site has averaged 32 days/year with daily average SST ≥ 28°C. As temperatures increase in the ocean, the predicted number of 28°C exceedance days also increases. Simulations of SST to the year 2100 give a mean of 18 and 37 exceedance days for the COC-OI and COC-Meas datasets, respectively ( Table S6 ). Simulations for the coastal bay site predict an average of 71 days/yr with SST ≥ 28°C in 2100 and a 5% chance that there could be 100 exceedance days in a year ( Figure 5 ). This would place the coastal bays in thermal stress conditions for most of the summer, with limited relief from lower temperature ocean water exchange through the tidal inlets. In a system otherwise hospitable to productive seagrass (Orth et al., 2020) and oyster communities (Smith et al., 2022), thermal stress associated with unrelenting high water temperatures may be the greatest threat to these ecosystems.

All annual MHW metrics are trending upward in the simulations to 2100 ( Figure 5 ; Table S6 ) primarily because of the positive trend in annual SST in all records ( Table 2 ). A MHW analysis of detrended SST time series at the coastal bay and ocean sites yields no significant trends in MHW metrics for the 40-yr records or the longer-term simulations. As discussed by Oliver et al. (2021), the choice of a fixed (as used here) vs. moving baseline analysis of MHWs depends on the questions motivating the study. If ecosystem vulnerability to heat stress is the main concern, especially when there appears to be a temperature tolerance threshold, then a fixed baseline analysis is valuable for considering the potential for future increases in temperature stress. In the simulations run here, the increase is greatest for MHW exposure and number of events, both of which increase about five-fold from 1982-2100. The MHW simulations did not explicitly account for connectivity between coastal bay and ocean water, although the record of MHWs from 1982-2021 for both sites in this study revealed a considerable degree of coupling between these systems, with high SST events generally occurring simultaneously at both sites even if they did not last long enough to qualify as a MHW ( Table S5 ). This coupling could affect future MHW characteristics in the coastal bays, particularly if trends in SST inside and outside the coastal bays are such that differences in seasonal climatology and threshold temperatures decrease with time. This reinforces the need for developing and maintaining long-term measurements of water temperature in shallow coastal systems in order to identify changes in conditions and trends as global ocean temperatures warm.

Publicly available datasets were analyzed in this study. This data can be found here: https://portal.edirepository.org/nis/mapbrowse?packageid=knb-lter-vcr.393.2, https://psl.noaa.gov/data/gridded/data.noaa.oisst.v2.highres.html, https://climate.esa.int/en/projects/sea-surface-temperature/, https://rucool.marine.rutgers.edu/integrated-ocean-technology/remote-sensing/, https://data.marine.copernicus.eu/product/SST_GLO_SST_L4_NRT_OBSERVATIONS_010_001/description, https://coralreefwatch.noaa.gov/product/5km/index_5km_sst.php, https://tidesandcurrents.noaa.gov/, https://www.ncei.noaa.gov/access/crn/, https://www.ndbc.noaa.gov/, https://portal.edirepository.org/nis/mapbrowse?scope=knb-lter-vcr&identifier=61. Code used to calculate temperature and marine heatwave statistics can be found here: https://github.com/plwiberg/MarineHeatwaveDataCode.

The author confirms being the sole contributor of this work and has approved it for publication.