The Pairing of Rapid Intensification Events and Eyewall Replacement Cycles in Tropical Cyclones in the Atlantic Basin from 2015 to 2020

Abstract

Rapid intensification (RI) and eyewall replacement cycles (ERCs) frequently occur in intense tropical cyclones (TCs), often causing rapid, significant changes in intensity and structure. In some TCs, RI and ERCs can occur concurrently or within a short period of one another. This study investigates whether there is a link between RI and ERCs by conducting a statistical analysis of TCs that occurred in the North Atlantic basin from 2015 to 2020. The HURDAT2 dataset was used to detect RI events, while the Morphed Integrated Microwave Imagery archive by the Cooperative Institute for Meteorological Satellite Studies was used to detect ERC events. Three sets of data were constructed from this analysis: TCs with paired RI/ERC events that occurred within 24 h of each other, TCs with RI only, and TCs with ERCs only. Statistics selected for analysis within the constructed datasets were mean duration of phenomena, mean rate of intensification, and mean peak intensity. We performed t-tests to determine the statistical significance of results. The results of this study show that TCs with these paired RI/ERC events often intensified at a faster rate, intensified for longer, and ended up stronger than TCs that only experienced RI or ERCs in isolation.

1. Introduction

1.1. Motivation

Intense tropical cyclones (TCs) like Hurricane Irma (2017) have undergone periods of rapid intensification (RI) and eyewall replacement cycles (ERCs), sometimes concurrently [1]. As often with both phenomena, large fluctuations in intensity may occur over a short period—often less than 24 h [2,3,4,5,6,7]. With more storms now forming in the Atlantic annually than before, it is important to understand how these two processes interact with one another [8]. However, the rate of RI has not seen a significant shift in climatological averages, as any change in RI may be disguised in other noise as well as its large variability over a timescale of years and decades [9]. The highly variable nature and behavior of RI make it difficult to track over long periods.

TC intensity and structure must be both parameterized and forecasted accurately, as errors can mislead the public into under-preparation, which can be costly. Struggles of forecasting RI in TCs with model guidance have been an area of focus to reduce error [10]. In the past decade, numerous improvements in model construction have been made to increase model accuracy [11,12,13,14]. This increase in model accuracy has also been concurrent with an improved understanding of RI and ERCs applied to modeling, as shown by the work of [3,15]. These improvements have translated to small (<10%) intensity forecast improvements up to 96 h [16]. Despite improvements in model guidance, there are still some limitations in intensity forecasting, as discussed in [16]. As shown in Figure 1, intensity error has decreased to 15 kt or less on all forecast timescales as of 2020 [17]. Despite this progression, there have been recent seasons (2015 and 2019) where the average intensity error has increased. Thus, the decrease in overall intensity error has not been uniform. Regardless, the current intensity error margin can still be high enough to allow a TC to strengthen or weaken to one Saffir Simpson Hurricane Wind Scale (SSHWS) category higher or lower than forecast [18]. Therefore, considerations need to be taken for both intensity and structure, as a better understanding of both can lead to better modeling and less uncertainty in TC forecasting.

Statistical guidance for TCs started with the introduction of the Hurricane Analog Technique (HURRAN) by [19]. The HURRAN was a model that selected a TC that happened in the past to forecast a developing TC’s track, plotting a probabilistic forecast. However, statistical intensity guidance was not developed until the creation of the Statistical Hurricane Intensity Prediction Scheme (SHIPS). The SHIPS is a model that was developed by [20] that takes the environment around TCs into consideration to predict intensity changes. These factors include wind shear, sea surface temperature, upper-level data, and a persistence bias to develop a model that considers synoptic and climatological predictors to forecast a TC’s intensity. The SHIPS is only one model out of a suite of models that the National Hurricane Center (NHC) uses for intensity guidance, with some models being a modified version of SHIPS [21]. The development of the SHIPS model and subsequent models has helped the NHC to reduce forecast intensity error in the Atlantic Basin, but the error margin is still large enough to underforecast a TC by one whole SSHWS category [17,18]. The relevance of this modeling overview to this study is that all of this past modeling work has contributed to the current human understanding of TC intensification processes—including RI and ERC—as well as our current forecasting capabilities that could be improved with this investigation.

1.2. Rapid Intensification

Rapid intensification is a process by which a hurricane has a 30 kt or more increase in winds within a 24 h timespan, as defined by [22]. In their case study range (1989–2000) in the Atlantic basin, [22] concluded that storms that were of a major hurricane intensity or stronger often underwent RI, with all storms in that timeframe that reached category four or stronger having at least one episode. Environmental conditions for RI to occur in the Atlantic include warm sea surface temperatures (often above 27 °C), high moisture in both the lower and middle troposphere, weak vertical wind shear, minimal upper-level forcing, and easterly upper-level flow [22]. Storms that failed to undergo RI did not have these environmental conditions and thus were discernible based on these factors within the study period. A subsequent study by [15] found that enhanced upper-level divergence also played a role in promoting RI, enhancing a TC’s outflow. For example, Hurricane Opal (1995) interacted with both a synoptic trough and jet stream which enhanced its outflow, causing unexpected RI in the Gulf of Mexico [23]. However, the warm core ring present in the Gulf of Mexico during Opal’s lifespan also played an integral part in Opal’s period of RI [24]. Opal’s RI event was also supported by model simulations that captured this jet stream enhancement [25]. Comparing environments with and without jet-stream-enhanced outflow, the numerical model findings of [25] support those of [15].

Rapid intensification has numerous triggering mechanisms. In their model simulations, ref. [26] found that RI was triggered when TCs became axisymmetric and formed an eyewall. With these structural improvements, TCs take advantage of the favorable conditions discussed in [15,22]. Furthermore, earlier work by [27] revealed that embedded mesovortices in a TC’s core can cause large, rapid pressure falls, given that an area of enhanced vorticity within the eyewall is present.

Ref. [28] proposed a new method of predicting RI, which looked at a ring pattern on 37 GHz microwave imagery that formed around the core of a TC as a predictor of RI. Their method performed well on events in the Atlantic from 2003 to 2007 when combined with the SHIPS Rapid Intensification Index (RII). Their method produced a 74% probability of RI (PRI), a 24% probability of detection (POD), and a 26% false alarm rate (FAR). Compared to the values of the 37 GHz ring pattern and SHIPS RII in isolation, the combined procedure had higher PRI, but lower POD and FAR values. These results mean that the blended method often detects fewer events, but, of these events, they were less likely to be false alarms.

1.3. Eyewall Replacement Cycles

Eyewall replacement cycles often present themselves in the form of concentric eyewalls, which are conspicuous on microwave (Figure 2), radar, and conventional satellite imagery [2,3]. Ref. [3] defines a secondary eyewall (SE) signature on microwave imagery as a partially closed ring outside of the primary eyewall, about three-quarters (75%) complete. For it to be classified as an SE formation event, the signature must occur within a 12 h timespan of a suspected start time of an ERC. This timeframe is to account for limitations with both satellite and in situ aircraft sampling, which are temporally inconsistent [3].

In general, ERCs occur in three distinct phases: intensification, weakening, and reintensification [2]. First, the storm intensifies while the eyewall and wind field contract. After strengthening to a peak, a secondary, convective ring forms outside a TC’s eyewall. This SE is identifiable by both the methods listed prior as a disconnected, circular secondary wind maximum that surrounds the main eyewall. The secondary ring then contracts around the main eyewall, causing it to weaken as the secondary eyewall becomes more dominant. The cycle completes when the original eyewall completely dissipates, leaving the SE as the new, primary eyewall. As a result of the structural changes that come with the formation of the larger eyewall, the wind field of the TC typically expands during an ERC event. It is only after the completion of an ERC that the wind field contracts again as the storm re-intensifies [4]. This larger storm structure can create devastating effects in the form of increased storm surge, larger rainfall storm totals, and a stronger storm after TC reintensification [4,29].

Sizable records on ERCs have been compiled, ranging from how long they take to how many have occurred, and storms that have multiple cycles have also been listed. Ref. [4] determined that a mean ERC takes 36.3 h to complete for a sample size of 24 events. ERCs typically occur in storms with 1 min sustained winds of 120 kt or greater [30], which is well within the major hurricane range of the SSHWS [18]. As such, their impact on intense cyclones resulting from RI is significant. Furthermore, about eight TCs exhibit one or more ERCs annually across multiple basins, including the North Atlantic (NATL) and the Eastern North Pacific (EPAC) [31]. When storms undergo multiple ERCs, 90% of TCs have larger primary eyewalls than their last ERC, and 80% have another ERC that takes longer than the prior ERC [4]. In fact, ref. [4] suggests that latter cycles are often more complex and larger in scope; thus, subsequent ERCs take longer.

For their study, ref. [3] constructed a database of SE formation events in the Northern Hemisphere from 1997 to 2006, specifically focusing on three basins: NATL, EPAC, and the central North Pacific. After running a complex probabilistic naïve Bayes model, they determined that SE formation in both the NATL and EPAC was more likely to occur in an environment conducive to the formation of deep convection, as supported by the environment having a high maximum potential intensity (MPI). For context, MPI is the calculated intensity ceiling of a TC that the environment can support [32]. Besides high MPI values, ref. [3] determined that other favorable environmental factors for SE formation include weak vertical wind shear, weak zonal flow at upper levels, high humidity through the middle and upper troposphere, and deep warm waters. Interestingly, some of these factors overlap with some favorable environmental conditions for RI, contributing to this study’s focus on these intertwined phenomena.

1.4. Paired RI/ERC Events

Paired RI/ERC events have been recently investigated in two storms: Hurricane Irma (2017) [1] and Typhoon Goni (2015) [33]. Ref. [1] noted that there were two ERCs while Irma underwent a period of RI, both of which appeared to have different mechanisms that formed the SE, as well as different progressions. The first cycle was typical in its progression, with the SE becoming increasingly dominant as the primary eyewall collapsed, whereas the second cycle saw the two eyewalls coalesce into one eyewall [1]. Ref. [1] also points out the brevity of both cycles, as they both took less than 12 h to complete—approximately one-third of the mean ERC of 36.3 h [4]. These results made it unclear how Irma kept undergoing RI amid back-to-back ERCs (Figure 3). Ref. [1] hypothesized that the contracting eyewall was paired with the advection of angular momentum surfaces. Regardless, ref. [1] felt that the intermittent sampling of Irma by the aircraft limited the scope of the data they needed to conduct a more proper analysis.

Typhoon Goni (2015) was another storm investigated for the close temporal proximity of RI and ERCs, as it underwent a period of RI after it completed an ERC. Using derived results from ground-based Doppler radar, ref. [33] used the ground-based velocity track display (GBVTD) technique to calculate the radius of maximum winds (RMWs) for Goni at multiple altitudes over 24 h. The GBVTD technique calculates the wind field at one level using velocity data from Doppler radar under the assumption that the TC has one defined circulation [34]. Ref. [33] used GBVTD to observe how Goni’s eyewall evolved while it underwent RI. As the eye contracted, the 1 and 5 km levels of the RMW contracted faster than the area between those altitudes. This intriguing contraction created a bulge in the RMW at the 3 km level. This initial contraction was paired with outflow greater than 2 m s⁻¹ observed above the boundary layer on both sides of the RMW, showing supergradient flow embedded in and above said boundary layer. The angle of the outward slope created by the contracting RMW peaked at approximately 45° before reducing to 38° at the end of the observational period. During most of the observation period, a radius of maximum reflectivity resided inside the RMW below an altitude of 9 km [33]. Using derived observations from the Doppler radar of Goni’s period of RI, ref. [33] suggests that an enhancement of boundary layer outflow after an ERC may lead to the contraction and intensification of a TC. However, ref. [33] also identifies that these suggestions are undermined by the limitations of the GBVTD technique and the absence of thermodynamic observations. Thus, the assertion that outflow from the boundary layer influences RI after an ERC, as seen with Goni by [33], is only hypothetical.

Prior paired RI/ERC studies have been on a case study basis, with studies on singular storms such as the ones conducted on Hurricane Irma (2017) by [1] and Typhoon Goni (2015) by [33]. Also, both case studies of paired RI/ERC seemingly asked more questions than they answered; thus, it is uncertain how both processes impact one another. No singular study has looked at overall trends with paired RI/ERC events over a defined period. Thus, this study proposes a method to investigate the characteristics of paired RI/ERC in intense TCs in a manner that has not been investigated previously.

2. Data and Methods

2.1. Tropical Cyclone Data

For the linkage of RI and ERCs, the 2015–2020 Atlantic Hurricane seasons were selected for the study. The short timeframe of the study period (6 years) was selected as an adequate scope for an exploratory introduction to a more in-depth investigation in the future (if merited). The Atlantic basin itself was chosen as a place of interest due to the recent uptick of activity noted by [8] and its potential for large, damaging storms that result from paired RI/ERC like Hurricane Irma [1,35]. For detecting RI, HURDAT2 data were used to detect events by the threshold of a 30 kt or greater increase in winds within 24 h [22,36]. For ERCs, the Morphed Integrated Microwave Imagery at CIMSS (MIMIC) archive from the Cooperative Institute for Meteorological Satellite Studies [37] was used in conjunction with the method outlined in [3] (75% or greater enclosure of the PE by the SE within 12 h of the suspected start time) to detect events.

A list of TCs that have an ERC and RI within 24 h of each other was constructed using the detection methods above, regardless of the order of occurrence. The order of RI/ERC events was neglected to allow us to account for cases like Irma and Goni, which both fell outside of the ERC paradigm proposed by [2]. The order of events was not a factor with this study, as the literature on paired RI/ERC events had no mention on whether the order was consequential [1,33]. The 24 h window is analogous to the mean 27 h it takes for the weakening and reintensification phases put forth by [4], and thus it was selected as the basis for the temporal threshold to define the events as paired.

As illustrated in

Table 1. A 5 × 5 table illustrating the selection process for adding events to the RI-Only, ERC-Only, and Paired RI/ERC lists.

Steps	Event A	Event B	Event C	Event D
1: >30 kt increase intensity since t = −24 h?	Yes	No	Yes	No
2: >75% enclosure of PE by SE on microwave imagery at t = 0 h?	No	Yes	Yes	No
If yes to (1) and (2), did both happen within 24 h?	N/A	N/A	Yes	N/A
List to be added to	RI-Only	ERC-Only	Paired RI/ERC	Neither

, each event is taken from a particular timestamp on the HURDAT2 dataset (i.e., use 20170831, 1200z for Event A—Hurricane Irma’s (2017) first RI period) [35]. A 6 h temporal resolution was selected for consistency, since most data within the HURDAT2 dataset omitting specific flags (i.e., landfall) had a temporal resolution of 6 h [36]. The data before the timestamp within 24 h were examined for RI in accordance with the threshold identified in [22]. If the period was less than 24 h, a scaled ratio was used to capture and record such instances (an example of this was Hurricane Chris (2018), which intensified 25 kt in 12 h—greater than the scaled rate of 15 kt/12 h). Then, the event was observed on CIMSS MIMIC imagery [37] around that timestamp for an SE eyewall signature as defined by [3], at a temporal resolution of 15 min. A check was completed to see if both events happened within 24 h of each other, regardless of order. After all these steps were completed, the event was categorized in a list depending on what criteria were satisfied. If no criteria were satisfied, the event was disregarded. After the completion of the steps outlined on the table, the next timestamp in the HURDAT2 dataset [36] was investigated in the same manner until the completion of a TC’s dataset. Analysis continued until the list of TCs between 2015 and 2020 in the NATL Basin was exhausted.

2.2. Methodology and Statistical Variables Selected for Analysis

The list of paired RI/ERC storms was compared to sets of null cases that included either RI or ERCs to see if there were discernable differences between the paired phenomena and each phenomenon in isolation. Characteristics that were investigated between the list of paired RI/ERC events and null sets were the mean duration of the phenomena, mean rate of intensification, and mean peak intensity. As discussed by [1] with Irma’s ERC times, the rate of ERCs in paired events may be much faster than the 36.3 h mean proposed by [4], and thus that characteristic was selected for ERC comparison. This mean, along with intensification rate and peak intensity means were selected to understand storm evolution with both phenomena concurrently compared to just one in isolation. Standard deviations (SDs) were calculated for all variables. One-tailed, unequal variance t-tests were performed for each mean value within the study to determine statistical significance.

3. Results and Conclusions

3.1. RI-Only Events vs. Paired RI/ERC Events

A list of 32 TCs that underwent RI (including 40 total RI events) was compiled from the 2015–2020 Atlantic Hurricane seasons using the HURDAT2 dataset. Of those 32 TCs, 21 underwent strictly RI and 11 underwent paired RI/ERC events. The mean value of intensification duration was slightly longer for paired RI/ERC cases (34.7 h) than strictly RI alone (29.5 h), illustrating that storms that undergo paired RI/ERC events intensify for longer. Intensification rates were quite comparable for both categories, with RI-only events (37.9 kt/24 h) having slightly faster intensification than paired RI/ERC events (38.3 kt/24 h). Regarding mean peak intensity, paired RI/ERC events were considerably stronger (125.0 kt) compared to strictly RI alone (101.9 kt), suggesting that storms that undergo both processes within a 24 h time frame intensify for longer and reach a stronger peak than storms that strictly undergo RI alone. These intriguing results do call into question whether the period of RI some TCs underwent triggered an ERC as described in the classic paradigm of ERCs in [2]. Is the temporal difference between RI and ERC events investigated here (within 24 h) the same process underlined in [2], albeit faster, or is there something more complex happening as discussed in [1,33]? The causality is unclear and is a notable point of focus to bring up for future investigations.

Standard deviations of mean intensification duration and rate of strictly RI events (8.0 h and 8.3 kt/24 h) were smaller than paired RI/ERC cases (12.8 h and 10.6 kt/24 h). However, the SD value of the peak intensity of RI-only events compared to paired RI/ERC events was larger (26.5 kt vs. 21.6 kt). These comparisons suggest that the intensification rate and duration data spread was tighter for RI-only cases and peak intensity data had a tighter spread for paired RI/ERC cases. The t-test value for the mean intensification rate was larger than 0.05 (t = 0.444), suggesting that these results (mean and SD) were statistically insignificant, meaning that they can be disregarded. The substantially large t-test value for this category is not unexpected, considering both intensification rate means were close to one another and were too close to discern any meaningful difference. Despite being larger than 0.05, the t-test value for intensification duration was still below the 0.10 threshold (t = 0.081), suggesting a marginally significant result. The t-test for peak intensity, however, was less than 0.01 (t = 0.007), suggesting a statistically significant result. These t-test results strongly suggest that paired RI/ERC events consistently yielded higher peak intensities than events that were strictly RI alone, and that paired RI/ERC events could have longer intensification periods than events that were strictly RI alone. A further analysis as to why paired RI/ERC TCs intensify and restructure in the manner that they do would be helpful as to why these storms behave the way they do. A logical place to start for this would be looking at synoptic scale conditions as highlighted by [3,22] for both RI events and ERCs to see if there is a specific overlap that would be clearer cut than the list of conditions for both phenomena by themselves. All calculated statistical values in this section are summarized in

Table 2. A table of statistics comparing RI-only cases compared to paired RI/ERC events.

Category (Units)	RI Only	Paired	t-Test (Unitless)	SD RI-Only	SD Paired
Number of TCs	21	11	N/A	N/A	N/A
Number of Events	24	16	N/A	N/A	N/A
Mean Intensification Duration (h)	29.50	34.69	0.081	7.96	12.82
Mean Intensification Rate (kt/24 h)	37.87	38.31	0.444	8.32	10.57
Mean Peak Intensity (kt)	101.90	125.00	0.007	26.53	21.56

.

3.2. ERC-Only Events vs. Paired RI/ERC Events

A list of 12 TCs that underwent ERCs (including 20 total ERC events) was compiled from the 2015–2020 Atlantic Hurricane seasons using the HURDAT2 dataset and MIMIC imagery. Of those twelve TCs, one only had an ERC, eight underwent paired RI/ERC events, and three had a mix of ERCs and paired RI/ERC events. The mean duration of the weakening phase for paired RI/ERC cases (16.6 h) was slightly shorter than the ERC-only counterparts (19.3 h). There was a stark difference in mean intensification rate for paired RI/ERC events (38.3 kt/24 h) versus ERC-only cases (3.0 kt/24 h). This difference illustrates that the paired events had significant intensification compared to ERC events alone, which had minimal to no intensification. Interestingly, ERC-only cases had a higher average peak intensity (127.5 kt) compared to the paired RI/ERC cases (125.0 kt). This could be attributed to the small sample size of the ERC-only dataset; however, it is in line with what [30] asserted about intense TCs (>120 kt) typically undergoing ERCs. Regardless, the results from this category have a degree of uncertainty from the small number of events and may need a larger sample size to obtain a more complete understanding of this comparison.

Despite the small sample size of the ERC-only dataset, there was a surprisingly narrow spread of mean weakening phase duration across the four events studied. With an SD of 6.2 h, the data spread of the ERC-only events was close in spread to the paired RI/ERC events (6.9 h), which had nearly triple the sample size of the events. However, even though the paired dataset had more incidents, the sample sizes of both datasets were still relatively small, and this could be considered a caveat of the analysis. Regardless, SD values for ERC-only cases were larger in mean intensification rate and mean peak intensity (20.2 kt/24 h, 26.6 kt) compared to the paired RI/ERC counterparts (10.6 kt/24 h, 21.6 kt), indicating higher variability and spread in the selected ERC-only case studies in these categories. The variable nature of the ERC-only dataset further highlights the need for a more extensive dataset for the mean values featured in this comparison. With these caveats in mind, the t-test analysis comparing the mean intensification rate was unsurprisingly statistically significant (t = 0.018, t < 0.05). Much in line with the paradigm set forth in [2], ERC-only cases saw middling intensity changes, whereas paired RI/ERC cases showed large changes in intensity, as previously noted in conventional RI studies (i.e., [22]). Other t-test results in this comparison for mean weakening phase duration (t = 0.242) and mean peak intensity (t = 0.437) suggest that these comparisons are not statistically significant. This result aligns with the small differences between mean weakening phase duration and mean peak intensity. Ideally, a larger sample size would be better at telling whether these comparisons could be safely disregarded, but, for the purposes of this study, a smaller sample size is acceptable. The small sample size could be used as potential motivation to perform a wider analysis over a larger time period with more cases in the future. All calculated statistical values in this section are summarized in

Table 3. A table of statistics comparing ERC-only cases to paired RI/ERC events.

Category (Units)	ERC Only	Paired	t-Test (Unitless)	SD ERC-Only	SD Paired
Number of TCs	4	11	N/A	N/A	N/A
Number of Events	4	16	N/A	N/A	N/A
Mean Weakening Phase Duration (h)	19.31	16.64	0.242	6.20	6.89
Mean Intensification Rate (kt/24 h)	3.00	38.31	0.018	20.23	10.57
Mean Peak Intensity (kt)	127.50	125.00	0.437	26.61	21.56

.

3.3. Conclusions and Future Work

Storms with paired RI/ERC events are stronger, intensify more quickly, and intensify for longer periods of time compared to storms that undergo RI or ERC events alone. Paired RI/ERC events intensify an average 23.1 kt more and intensify an average 5.2 h longer than storms that undergo RI alone. Paired RI/ERC events also intensify an average 35.3 kt/24 h faster than storms that undergo ERCs alone. However, this value is in line with the definition of RI of 30 kt/24 h denoted in [22]. These characteristics are attributed to the TC restructuring its core in a manner that does not impede intensification, often beginning to do so before intensification is over. Thus, the TC is further along or finished with the weakening phase of an ERC before it peaks in intensity, at the cost of taking slightly longer to intensify due to the underlying structural changes. The manner of paired RI/ERC phenomena in TCs handling both core restructuring and intensification makes them dangerous and important to study further.

However, there are underlying problems with the results of this study, as highlighted by the small sample size of the case study pool. An expanded study with TCs from multiple basins across an expanded timeline could produce more robust, complete results. However, for an introductory study of both phenomena to merit further investigation, this study satisfied that basis to motivate further work with this methodology. There are limitations to how far back the study period can go as CIMSS MIMIC data only goes back to 1999 at its absolute earliest prototyping in the NATL basin with Hurricane Floyd [37]. No complete CIMSS MIMIC dataset exists before 2005 [37], which is the most reasonable limit for this study’s evolution.

For further work investigating how these two phenomena interact with another within a 24 h time window, a reasonable starting point would be to further expand the study period closer to its absolute limit in the NATL basin, as well as include new data from recent hurricane seasons in the NATL basin later than 2020. A completion of this dataset in the NATL basin would create a strong foundation for further work in other basins around the world. The next logical step after that would be including data from the EPAC basin followed by data from the Northwestern Pacific basin. With the frequency of TC formation in the Northwestern Pacific Basin in an average season (26.7 tropical storms and 16.8 hurricane-strength typhoons) [38], including this dataset along with the EPAC dataset (15.4 tropical storms, 8.3 hurricanes and 4.2 major hurricanes on average annually) [39] could greatly increase the depth and scope of this study.

Not only would an expanded sample size assist in the future, but a better understanding of causality within the paired RI/ERC events (i.e., did something within the RI event trigger the ERC?) would greatly increase the depth and clarity of the analysis. Furthermore, examining synoptic scale conditions that specifically allow for paired RI/ERC events would also help to bolster results in tandem with the larger sample size. This study’s results have laid the groundwork for future analysis regarding paired RI/ERC events. With future studies, the behavior of paired RI/ERC events can be understood beyond this study’s initial look at paired RI/ERC events. In particular, further work will lead to a better understanding of their faster intensification rate, strengthening duration, and higher peak intensity compared to their non-paired RI/ERC counterparts.