Tropical cyclones (TCs) are one of the costliest and deadliest natural disasters on the planet (Smith, 2020). The ability to forecast TC intensity has improved recently (Cangialosi et al., 2020), however intensity forecast errors remain large (~12 kts at 72 hours). The primary controls of the intensity of mature TCs are vertical wind shear, dry air intrusion, and the fluxes of enthalpy and momentum between the surface ocean and atmosphere (Emanuel, 1986). Numerous studies have shown that the upper ocean can evolve rapidly beneath TCs and feedback on storm intensity (Cione and Uhlhorn, 2003; Black et al., 2007; D’Asaro et al., 2007; Zedler et al., 2009; Mrvaljevic et al., 2013; Steffen and Bourassa, 2020) among many others therein. Recent studies have focused on coastal ocean processes and their feedbacks on storm intensity, including coastal upwelling, downwelling, and enhanced shear-driven mixing (Glenn et al., 2016; Seroka et al., 2016; Miles et al., 2017; Seroka et al., 2017; Zhang et al., 2018; Dzwonkowski et al., 2021; Gramer et al., 2022), with a particular focus on highly stratified water columns. A common hurricane intensity forecasting challenge in regions with large river runoff are upper ocean salinity barrier layers (Lukas and Lindstrom, 1991; Sprintall and Tomczak, 1992; Foltz and McPhaden, 2009; Grodsky et al., 2012; Steffen and Bourassa, 2018). Generally, stratification can inhibit vertical mixing and limit entrainment of cool subsurface waters into the mixed layer during TC passage (e.g. Rudzin et al., 2018). These ocean features are found on continental shelves, near river outflows (Sengupta et al., 2008), and over the open ocean with offshore transport of freshwater (Pailler et al., 1999; Grodsky et al., 2012). Barrier layers increase the potential energy gradient, inhibit sea surface temperature (SST) cooling, and can support enhanced enthalpy fluxes into the atmosphere during hurricanes (Wang et al., 2011; Balaguru et al., 2012; Rudzin et al., 2018; Rudzin et al., 2019; Balaguru et al., 2020; Rudzin et al., 2020; Sanabia and Jayne, 2020). Only a few observations and studies have explicitly focused on the interactions of TCs passing over the Mississippi river-induced salinity barrier layer (Le Hénaff et al., 2021; John et al., 2023). This barrier layer is a product of the largest river outflow in the US from the Mississippi River, in a region where strong hurricanes make landfall and coastal communities have repeatedly been devastated by powerful landfalling hurricanes, including Hurricane Ida in the summer of 2021.

Hurricane Ida (2021) underwent rapid intensification (RI) over the warm waters of the Gulf of Mexico (GoM) ( Figure 1 ), with an increase in maximum wind speed of 60 kts (~30 m/s) in 24 hours (Beven et al., 2022). Ida continued to intensify as it passed over the continental shelf before making landfall as a category 4 hurricane in Louisiana on August 29^th ( Figure 1 ) as the second costliest storm to make landfall in the region after Hurricane Katrina (2005); (Smith, 2020). A recent study (Zhu et al., 2022) identified that nearshore SSTs ahead of Ida were >30°C, above the mean SSTs (28.7°C) that other major hurricanes crossed over in the region. They also indicated that slow translation speeds kept the backside of the storm over these warm and fresh waters for an extended duration, contributing to Ida’s slow weakening after landfall. While there is a large body of research on freshwater plume, or salinity barrier layer, impacts on hurricane intensity there are only a few focused on the Mississippi River plume (Le Hénaff et al., 2021; John et al., 2023). Despite major hurricanes regularly transiting this region, there are limited upper ocean observations during storm events in this region. For example, in the highly dynamic region where Ida rapidly reached and maintained category 4 ( Figure 1 ) from 27.5° to 30° N and between 91° and 88.5° W only 20 Argo floats and 252 profiles are available in the last 20 years (~13 profiles/year) during hurricane season (https://erddap.ifremer.fr/erddap/index.html).

According to Beven et al. (2022), official forecasts for Hurricane Ida (2021) generally outperformed guidance and the previous five year mean official forecasts for the full storm period. However, few models or official forecasts captured Ida’s peak winds at landfall including as Ida rapidly intensified over the warm waters of the central GoM and fresh Mississippi River plume coastal waters ( Figures 2 , 3 ). Fortunately, as part of the 2021 Hurricane Glider Program (Miles et al., 2021) a Navy operated and NOAA coordinated autonomous underwater glider, NG645, was deployed ahead of and during Ida’s eye passage over the region ( Figures 1 – 3 ). Ahead of the storm, in the deep ocean (>100m depth) just south of the GoM northern escarpment NG645 observed ( Figure 2 ) warm sea surface temperatures, low salinity, and heat content near a threshold (60 kJ cm^-2) typically conducive for intensification (Mainelli et al., 2008). The presence of the freshwater barrier layer and elevated SSTs suggest that, even with marginal ocean heat content, these ocean conditions are conducive to storm intensification. In this study we investigate upper ocean metrics for storm intensification in the region Hurricane Ida (2021) passed over, as well as the sensitivity of SST cooling to the strong vertical salinity stratification in the region ahead of landfall. To carry out this work we combine the in-situ observations from NG645 and satellite remote sensing with a 1-D mixed layer model sensitivity experiments to evaluate the impact of barrier layer presence and absence with the Price-Weller-Pinkel (PWP) model (Price et al., 1986).

Ocean observations ahead of and during Hurricane Ida were obtained from Slocum glider (Schofield et al., 2007) NG645, operated by the Naval Oceanographic office in close collaboration with the Integrated Ocean Observing System (IOOS) Hurricane Glider Program. Slocum gliders are buoyancy driven uncrewed underwater vehicles that can profile vertically (up to 1000 m at ~20 cm s^-1) and horizontally (~ 20 km day^-1). They typically collect data at up-to 2 second intervals, resulting in high (<1m) vertical resolution. These systems have been used over the past decade to study upper ocean processes during TCs (Domingues et al., 2015; Glenn et al., 2016; Seroka et al., 2016; Miles et al., 2017; Seroka et al., 2017; Lim et al., 2020) and to provide near real-time data for assimilation into operational hurricane forecast models (Miles et al., 2021). NG645 specifically was operated as part of an agreement between NOAA and the Navy with the goal of providing real-time in-situ glider observations to improve and inform hurricane intensity forecasts.

NG645 was deployed on June 13^th, 2021, offshore of the continental shelf at 27.6°N and 94.6°W. In mid-August the glider was navigated eastward south of the escarpment of the northern GoM ( Figures 1 – 3 ), through the northwestern edge of a loop current eddy (LCE) and into a gap region south of the continental shelf, but north of the LCE. The glider transited in the deep (>1000 m) of water off the continental shelf for the duration of the mission. NG645 crossed ahead of Ida’s track at 89.23°W and 28.12°N on August 19^th, 60 km from the shelf-break and 100 km from the nearest land point. The glider did not station keep at this location but was piloted to collect a broad swath of data further eastward before station keeping on August 27^th ahead of the storm at 88.17°W and 28.57°N. The region to the east of Ida’s track was a gap region between the continental shelf to the north and the LCE to the south. In this study we present data from NG645 through August 31^st, however the glider continued sampling through September 24^th in further support of hurricane forecast models. NG645 was equipped with a standard Seabird Scientific, Inc. (SBE) pumped conductivity, temperature, and depth sensor (CTD), which reported data in at ~8s intervals. The Naval Oceanographic Office submitted data in near real-time and for archiving via the IOOS Glider Data Assembly Center (DAC). However, post-deployment data was not made available, thus intermittent data transmission issues resulted in periodic data gaps.

Hurricane Ida impacted Cuba and entered the GoM late on 08/27 and into 08/28 as a category 1 storm ( Figure 1 ). It began to intensify over the central GoM late on 08/28, and rapidly intensified to a category 4 storm over the northern GoM and continental shelf until landfall at 08/29 16:00, gradually weakening on 08/30 as it moved inland ( Figure 1 ). A zoomed in view of Ida’s track and intensity in relation to glider NG645’s position and pre-storm in-situ SST, salinity, and OHC are shown in Figure 2 . These in-situ data show that the upper ocean was warm, and a freshwater barrier layer was present to the right of the storm track in the week prior to Ida’s passage.

SST imagery ahead of the storm (08/25) showed warm pre-storm SSTs above 30°C along the storm track ( Figure 3 ). The first clear composite image was available approximately 4-days after landfall and showed significant cooling ( Figures 3 ) of more than 1°C over more than 230,000 km² of the GoM, a maximum cooling of 3.8°C on the shelf near the landfall location and 2.38°C near the glider station keeping location. Ida’s rapid intensification despite this cooling implies that a significant portion of the satellite observed SST cooling occurred after the storm’s eye-passage. We use in-situ glider data to investigate the specific timing of the cooling further.

Cross-sections ( Figure 4 ) and derived upper ocean metrics ( Figure 4 ) demonstrate pre-storm ocean properties during the cross-track storm survey period (08/17 – 08/27) and the ocean response to the right of the storm track during the station keeping period (08/27 – 08/31). During the pre-storm survey period the glider observed an isothermal warm (>30°C) layer extending to ~30 meters depth to the west and ahead of the storm track. As the glider progressed eastward the isothermal layer shoaled to<20 meters. In contrast, the MLD was found near the surface (<5m) because of a shallow layer of low salinity water (~ 32.5 to 34.5 PSU) aside from a brief salty surface salinity on 08/17.

T100 showed warm average upper ocean temperatures to the west peaking at (28.2°C) and cooler temperatures to the east reaching a minimum of 23.4°C where the glider began its station keeping mission ( Figure 5C ). Ocean heat content ( Figure 5D ) had a similar pattern as T100, notably with values above the 60 kJ cm^-2 threshold identified for hurricane intensification by Mainelli et al. (2008) on the western portion of the track. Observed OHC dropped below that threshold on 08/23 as the glider progressed eastward reaching a minimum of 25 kJ cm^-2 as the glider started its station keeping mission. Aside from a brief dip to 400 J m^-3 on 08/21 the PEA remained near 500 J m^-3 throughout the pre-storm survey ( Figure 5E ). The consistently high PEA indicates that the water column stability was high, and the SST was not likely to cool to the full T100 value, despite the strength of the storm. For context, later in section 4 we detail the difference in PEA with and without a barrier layer as shown in ( Figure 6 ).

During the station keeping period the glider showed that the upper ocean cooled, increased in salinity, and both the ILD and MLD deepened throughout and following the storm event ( Figures 5F–J ). From 8/28 to eye-passage and landfall the SST cooled by 1.1°C and 1.96°C, respectively. This represents less than half of the satellite observed ocean cooling by eye-passage, and 82% by landfall. Sea surface salinity only experienced a small increase of 0.44 PSU for a brief period between eye-passage and landfall ( Figure 5G ). The MLD and ILD deepened from ~5m to ~20m and ~18m to ~30m. T100 and OHC experienced negligible changes throughout the storm mixing period ( Figures 5H, I ), while PEA dropped almost 100 J m^-3 from the station keeping period to landfall and continued to drop to 350 J m^-3 following landfall ( Figure 5J ). The minimal changes in T100 and OHC paired with a large drop in PEA suggest ocean mixing processes were first breaking down the upper ocean salinity stratification before accessing deeper cold subsurface waters. We evaluate this with the 1-D PWP model in the following section.

Observations from Slocum glider NG645 ahead of and beneath Hurricane Ida (2021) in the GoM captured upper ocean cooling ahead of eye-passage and landfall ( Figures 4 , 5 ). Despite this cooling, SST at the glider location just prior to landfall remained warm (28.1°C), approximately 3.5°C above T100 (24.6°C) at landfall. This indicates that the standard assumption made in Price (2009) that a typical category 3 hurricane will mix to ~100m was not valid for Hurricane Ida (2021) passing over the northern GoM. The freshwater barrier layer located over the deep ocean suggests that the northern GoM could be added to the list of regions where T100 is an unreliable metric such as the Bay of Bengal (McPhaden et al., 2009) or western Tropical Pacific (Price, 2009). Additionally, the glider observed OHC was just above the intensification threshold of 60 kJ cm^-2 suggested by Mainelli et al. (2008) in its pre-storm survey period ( Figure 6 ) and well below that threshold during the glider station keep period. Despite this relatively low OHC, Ida underwent RI and maintained its status as a Category 4 storm as it passed over the glider sampled region and made landfall. This indicates that OHC in this region was a poor metric for storm intensification, again likely due to salinity stratification as described in Price (2009). In contrast, both in the pre-storm survey and glider station- keep time periods ( Figure 5 ) PEA of the upper 100m suggested the water column was highly stable. For reference, the PEA of the initial profiles used in PWP Exp1A and Exp2A were 460 and 477 J m^-3. With the barrier layer removed in Exp1B and Exp2B the initial PEAs were reduced to 327 and 381 J m^-3, respectively. This represents a reduction in stability of 29% and 20%, respectively with the largest reduction in the along-track region.

The reduced cooling in PWP experiments at both the along-track and glider locations simulated here due to the barrier layer is consistent with previous studies focused on other regions. In these twin model experiments the salinity barrier layer inhibited SST cooling in Exp1 (and Exp2) by 53% (57%) ahead-of-eye passage, 38% (32%) by landfall. For example, Balaguru et al. (2012) identified a 33% reduction in cooling due to barrier layers in the category 4 hurricane Omar (2008) in the northeastern Caribbean. In the Bay of Bengal Neetu et al. (2012) showed that monsoon generated barrier layers are responsible for a ~40% reduction in cooling by TCs relative to post monsoon seasons. Idealized PWP experiments in Rudzin et al. (2018) were designed to represent a range of ocean features in the eastern Caribbean that showed cooling ranges of 0.4 to 0.8 °C, also consistent with the total cooling presented here. Similarly, an idealized coupled numerical modeling barrier layer sensitivity study (Hlywiak and Nolan, 2019) showed reduced cooling of more than 0.6°C for TCs that were slow moving, strong, and with favorable atmospheric conditions for generation using barrier layer conditions typical of the Amazon-Orinoco River Plume. Balaguru et al. (2020) also carried out extensive PWP model experiments with and without salinity stratification for the Amazon-Orinoco River plume to evaluate the connection between rapid intensification (RI) and salinity barrier layer cooling inhibition. They found for idealized RI cases, salinity barrier layers reduced SST cooling by up to 0.3°C, which is consistent with what we simulated for a rapidly intensifying Hurricane Ida. For non-RI cases in their study, salinity barrier layers were only responsible for inhibiting 0.15°C of cooling, highlighting potential feedbacks between barrier layers and RI. Despite their findings in the Amazon-Orinoco River Plume, they found that barrier layers had limited impact on storm intensity in the GoM. Their study utilized the Navy Global Ocean Forecast System (GOFS) to initialize PWP. The dearth of upper ocean observations to support data assimilation in the northern GoM, and practice of using climatological river inputs in GOFS may have limited their ability to resolve the sharp upper ocean salinity gradients such as those observed by NG645.

One of the few studies (Le Hénaff et al., 2021) of barrier layer and hurricane interactions in the GoM identified a barrier layer ahead of Hurricane Michael (2018). They identified SSS<34 PSU to as far south as 27.5°N, above the 32.6 PSU SSS observed in the pre-storm survey by NG645 ( Figure 2 ) but ~75km further south. However, a study of the intensification of Hurricane Isaac (Jaimes et al., 2016), which followed a similar track to Ida, found no evidence of barrier layers in profiles collected from air-deployed expendables. A recently published study of the ocean conditions ahead of Hurricane Sally (2020) (John et al., 2023) identified similar freshwater salinity barrier layers from the Mississippi River Plume as observed here in Ida, which contributed to continued intensification over the continental shelf. A study investigated the evolution of barrier layers during TCs globally with Argo floats (Steffen and Bourassa, 2018) using a barrier layer potential energy (BLPE) metric with similarities to PEA. They showed Argo floats between 2001 and 2014 with both low BLPE approaching 0 J m^-2 and high >1200 J m^-2 near overlapping at our study site. These studies and our findings indicate that barrier layers in the GoM are highly variable and can cover broad areas that hurricanes, such as Isaac (2012), Michael (2018), Sally (2020), and Ida (2021) must pass over before making landfall, and can have an impact on intensity.

The observations from NG645 and the results from the PWP sensitivity study highlight the potential importance of salinity stratification on the deep open ocean region off the continental shelf in the northern GoM, which is clearly influenced by coastal freshwater inputs. In the northern GoM a variety of thermal stratification regimes exist. In the nearshore environment it can be warm throughout the water column or highly thermally stratified. Ahead of Hurricane Michael (2018) subsurface temperatures of 22°C were observed near 10m depth to the northeast of our study-site (Dzwonkowski et al., 2020). This feature was removed during a marine heatwave that dramatically warmed the shelf waters to over 28°C in a few days. Further offshore the presence of LC/LCE waters can lead to warm and salty features that extend throughout the upper 100m and beyond (Elliott, 1982). In contrast, the region between the LC/LCE and continental shelf is typically warm at the sea surface but can have cooler waters beneath the seasonal thermocline in the upper ocean, as evidenced in the glider observations by NG645. In this “gap” region between the LC/LCEs and the continental shelf where the Mississippi River plume can be exported off the continental shelf, our findings suggest that salinity barrier layers can increase stratification and further isolate the subsurface cold water from mixing and cooling the surface. These warm and fresh surface waters would theoretically support storm intensification approaching landfall, or at a minimum reduce the oceans contribution to storm weakening.

We have shown that the standard upper ocean metrics OHC and T100 were likely not robust indicators of storm intensity in the deep waters of the northern GoM escarpment ahead of Ida’s landfall. An alternative stability metric, PEA, and 1-D upper ocean mixing model experiments indicated that the presence of a freshwater barrier layer likely inhibited additional sea surface cooling and enhanced enthalpy flux under a rapidly intensifying Hurricane Ida (2021). In our experiments the removal of the barrier led to earlier, more rapid, and greater cooling, which resulted in reduced enthalpy flux to the atmosphere, and a greater DPI. This is particularly critical as it highlights an essential ocean feature, a Mississippi River plume freshwater barrier layer, in the “gap” region south of the continental shelf and north of the LC/LCE that landfalling hurricanes must cross before impacting coastal communities. While the limited utility of OHC on continental shelves and T100 in freshwater stratified layers is well-known (McPhaden et al., 2009; Price, 2009; Potter et al., 2019), salinity observations are severely lacking in this region. This study highlights the need and capability of expanded ocean observing assets along the shelf-break of the GoM to identify freshwater barrier layers and improve intensity forecasts of landfalling hurricanes in this vulnerable region.

The datasets presented in this study can be found in online repositories. The names of the repository/repositories and accession number(s) can be found below: NG645 was operated by the Naval Oceanographic Office and coordinated by the Integrated Ocean Observing System (IOOS) Regional Association Gulf of Mexico Coastal Ocean Observing System through the Commercial Engagement Through Ocean Technology Act of 2018. Data sources included glider NG645 (https://gliders.ioos.us/erddap/tabledap/ng645-20210613T0000.html), Buoy 42040 from the National Data Buoy Center (https://www.ndbc.noaa.gov/station_history.php?station=42040). The PWP model was adapted from https://github.com/earlew/pwp_python_00.

TM, SC, and SG designed the initial experiments and carried out analysis. JE generated figures and carried out data analysis. SC, SG, JR, and ST provided scientific guidance and paper edits. TM prepared the manuscript for publication. All authors contributed to the article and approved the submitted version.