The NOAA Gulfstream IV (G-IV) routinely deploys global positioning system dropwindsondes (GPS sondes) to sample the environment around hurricanes that threaten landfall in the United States and neighboring countries. Part of this G-IV synoptic surveillance flight pattern is a circumnavigation 300–350 km from the circulation center of the hurricane. Here, the GPS sondes deployed over two consecutive days around Hurricane Felicia (2009) as it approached Hawaii are examined. The circumnavigations captured only the final stages of decay of the once-category-4 hurricane. Satellite images revealed a rapid collapse of the deep convection in the eyewall region and the appearance of the low-level circulation center over ~8 h. Midlevel dry air associated with the Pacific high was present along portions of the circumnavigation but did not reach the eyewall region during the period of rapid dissipation of the deep clouds. In contrast, the subtropical jet stream (STJ) enhanced the deep-layer vertical shear of the horizontal wind (VWS; 850–200 hPa) to greater than 30 m s−1 first in the northwest quadrant; ~6 h later the STJ was estimated to reach the eyewall region of the hurricane and was nearly coincident with the dissipation of deep convection in the core of Felicia. Felicia’s demise is an example of the STJ enhancing the VWS and inhibiting intense hurricanes from making landfall in Hawaii. The authors speculate that VWS calculated over quadrants rather than entire annuli around a hurricane may be more appropriate for forecasting intensity change.
Accurate forecasting of the decay of an overwater tropical cyclone (TC) has obvious benefits to marine interests and coastal communities expecting landfall. The study of TC decay, however, has taken a distant second place to intensification. Hurricane Opal (1995) stands as one of the few TCs where decay has been explored prior to extratropical transition or landfall (e.g., Rodgers et al. 1998; Bosart et al. 2000). Two eastern Pacific hurricanes, Jimena (1991) and Olivia (1994), are examples of how increasing vertical shear of the horizontal wind (VWS) can impact intensity (Black et al. 2002).
In August 2009, TC Felicia deepened to category 4 as it approached the Hawaiian Islands. This threat prompted the National Oceanic and Atmospheric Administration (NOAA) and the Central Pacific Hurricane Center (CPHC) to mobilize the NOAA Gulfstream IV (G-IV) for synoptic reconnaissance. Missions similar to these have reduced track forecast errors in the Atlantic basin (Burpee et al. 1996), now by as much as 32% (Aberson and Franklin 1999). For Felicia, we wish to explore if the global positioning system dropwindsondes (GPS sondes) deployed from the G-IV can provide insight into this TC’s rapid filling close to Hawaii. Specifically, we will examine the interaction of the subtropical jet stream (STJ) with the upper-level circulation of the TC.
Hawaii is in an enviable location during the hurricane season because the tropical upper-tropospheric trough (TUTT) is fully developed and its axis lies just north of the islands (Sadler 1975). The TUTT displaces the STJ southward over the archipelago. Given the underlying trade wind flow, the result is a belt of strong VWS over the islands (Fig. 1). Strong VWS has been implicated in the weakening of TCs through several physical mechanisms (e.g., McBride and Zehr 1981; Zehr 1992; DeMaria 1996; Bender 1997; Frank and Ritchie 2001; Gallina and Velden 2002; Tang and Emanuel 2012; Dolling and Barnes 2014) and currently serves as a predictor in the Statistical Hurricane Intensity Prediction Scheme (SHIPS; DeMaria et al. 2005).
2. Data and methodology
On 8 and 9 August the G-IV sampled the synoptic environment to the west-northwest of the TC, and then performed a circumnavigation of Felicia at a distance approximately 300–350 km from the circulation center (Figs. 2a,b). This distance is partially driven by studies by Gray (1989, his appendix) and Franklin (1990) that revealed TC motion is best correlated with conditions in that radial range. Over 90% of the GPS sondes were successful; the sonde deployment locations including the failures are depicted in Fig. 2. The period of circumnavigation was from 0840:55 to 1059:05 UTC 8 August and from 0919:40 to 1137:05 UTC 9 August.
The GPS sonde performance is described by Hock and Franklin (1999). The sondes have a 2-Hz sampling rate that translates to vertical resolutions of ~12–14 m at 200–300 hPa and ~5–7 m in the lower troposphere. All sondes were processed with the Atmospheric Sounding Processing Environment program (Martin 2007), then individually scrutinized to eliminate errors such as sensor wetting, following Bogner et al. (2000). The GPS sondes are processed using the methodology developed by Dolling (2010), and reviewed by Dolling and Barnes (2014). All data are storm relative with TC motion subtracted from Earth-relative winds and the sonde location is relative to the TC. During each mission Felicia (2009) filled by only a few hectopascals, so we assume a steady state during each circumnavigation.
Other datasets that we utilized included 1) the Optimum Interpolation Sea Surface Temperature, version 2 (OISSTv2), fields generated by NOAA for the week of 5–12 August; 2) the infrared, visible, and water vapor images from the Geostationary Operational Environmental Satellite (GOES); 3) estimates of TC position, intensity, and minimum central pressure from the NHC best-track (BT; Jarvinen et al. 1984) dataset; 4) tropical cyclone products and estimates of intensity from the Cooperative Institute for Meteorological Satellite Studies (CIMSS) at the University of Wisconsin–Madison [CIMSS uses the advanced Dvorak technique (ADT; Olander and Velden 2007) to estimate intensity]; 5) estimates of maximum potential intensity (MPI; Emanuel 1986); 6) National Centers for Environmental Prediction–National Center for Atmospheric Research (NCEP–NCAR) reanalyses with 2.5°-latitude resolution (Kalnay et al. 1996); and 7) forecasted intensity from the SHIPS dataset (DeMaria et al. 2005). A comprehensive discussion of the datasets and techniques may be found in Bukunt (2014).
a. Felicia’s track, initial weakening, and brief reintensification
Felicia achieved tropical depression status at 1800 UTC 3 August and became a tropical storm at 0000 UTC 4 August when it was located at approximately 12°N, 122°W (Kimberlain et al. 2010). With SSTs from 28° to 29°C (Fig. 3) and a synoptic regime characterized by low VWS, high total precipitable water (>55 mm), and a moist midlevel troposphere, Felicia rapidly intensified from 0000 UTC 4 August through 0000 UTC 6 August while being steered to the west-northwest by a well-defined deep-layer ridge. Felicia achieved a minimum central pressure of 935 hPa at 0000 UTC 6 August 2009.
The estimated minimum sea level pressures of Felicia based on BT data, the ADT, and MPI are shown in Fig. 4. For the MPI estimate we have used SST − 1.2°C as the proxy for inflow temperature, following the results of Cione et al. (2000), who determined the average difference between sea and air. Our estimate is derived from their data for the annulus from 0.5°, the approximate eyewall location, to 4.0° radius from the center. Felicia came within 5 hPa of its MPI. During this time BT data, the ADT, and MPI exhibited an equivalent weakening trend from 0600 UTC 6 August to 0600 UTC 7 August. This initial decay primarily reflects the reduction in inflow temperature as Felicia tracked over SSTs that decreased about 2°C (Fig. 3; note nearly parallel lines of MPI and actual intensity estimates).
From 0600 UTC 6 August to 0600 UTC 7 August there are no G-IV observations around Felicia, so we have to rely on the NCEP–NCAR reanalyses fields and CIMSS tropical cyclone products to infer environmental conditions. The CIMSS deep-layer VWS (850–200 hPa; not shown) remained less than 10 m s−1 within 500 km of the hurricane center. Total precipitable water was steady at 55–60 mm within that same radial distance. Temperatures in the 200- and 250-hPa levels varied only about 1°C during this initial decay period. Based on satellite images and the reanalyses upper-level wind fields, there was one strong outflow channel to the northwest and weaker flow to the southwest. Throughout the period, upper-level divergence did not vary by more than 15%. Other than the passage over cooler water, there were no apparent changes in other environmental factors that would be strong candidates for causing the initial decay.
The ADT shows a reintensification from 1800 UTC 7 August to 0000 UTC 8 August, again showing Felicia approaching its MPI. This was when the area of cold cloud tops over the circulation center expanded considerably. After 1200 UTC 8 August both BT data and the ADT rapidly diverge from MPI, suggesting that this phase of weakening, in contrast to the initial weakening, may not be readily predicted based solely on theoretical thermodynamic controls. Based on MPI theory, Felicia should have been nearly steady at 952 hPa after 1200 UTC 8 August. MPI does not take into account VWS or any internal TC processes such as eyewall replacement cycles or competing rainbands. The theory assumes that the TC has come into thermodynamic equilibrium with a maximum inflow temperature, a minimum outflow temperature, and a relative humidity in the environmental inflow of 80%. There is no accounting for the time it would take a TC to reach its MPI.
b. Evidence of the STJ’s intrusion into the eyewall
The storm-relative winds derived from the GPS sondes deployed during the two G-IV missions allow us to estimate when the STJ reached the eyewall of the TC. Along the circumnavigation on 8 August, 200-hPa winds were ~5 m s−1, with anticyclonic flow along the northwest–east portions of the ring and cyclonic flow to the southwest and south (Fig. 5a). In contrast, the winds on 9 August had a westerly component throughout the circumnavigation, with over 15 m s−1 westerly winds on the west side of the TC (Fig. 5b). Overall, the storm-relative flow at 200 hPa illustrates the intrusion of westerly flow associated with the STJ; the west side of the circumnavigation had strong west winds while more variable winds, in both magnitude and direction, characterized the eastern half. However, the winds still exhibited a westerly component on the east side of the circumnavigation, supporting our conjecture that the STJ’s contribution to adverse VWS had reached the eyewall of Felicia before the sampling on 9 August.
The storm-relative winds at 250 hPa demonstrated a wavenumber-1 asymmetry and an evolving interaction between the STJ and the TC. On 8 August, tangential flow (Fig. 6a) was weak throughout the majority of the circumnavigation save for the southwest quadrant supporting the contention that the circumnavigation was close to the inflection point where flow turns from cyclonic to anticyclonic. Tangential flow on 9 August becomes more asymmetrical, with stronger cyclonic flow in the southwest rapidly shifting to anticyclonic flow along the northwest through northeast portions of the circumnavigation because of the STJ. This pattern is evidence that a zone of diffluence existed on the west side of the circumnavigation. The storm-relative radial flow at 250 hPa reveals a remarkable intrusion of the STJ (Fig. 6b). Specifically, 8 August was characterized by radial inflow from −3 to −4 m s−1 from the southwest through northwest parts of the circumnavigation while on 9 August the inflow exceeded −20 m s−1 in the west-northwest sector. While significant inflow is evident to the west and northwest, outflow characterizes the opposite side of the TC.
Since the amount of moisture at 250 hPa is minimal, analysis of equivalent potential temperature θe is comparable to potential temperature θ. The θe values along the circumnavigation at 250 hPa indicate the intrusion of the STJ was not accompanied by a substantial change in θe, or θ (Fig. 6c). As shown, θe on both 8 and 9 August only varied from 345 to 348 K. The intrusion of the jet along the northwest quadrant was associated with an increase in θe of only 1–1.5 K.
The VWS from 225 to 850 hPa exemplifies the adverse impacts of the STJ (Fig. 6d). Here, we have calculated the shear from two layers (200–250 and 825–875 hPa) to avoid any unrepresentative observations from a given level. There were few observations above 200 hPa precluding the use of the 175–225-hPa layer. Along the 3° latitude circumnavigation, 8 August exhibited wind shear of 8–15 m s−1 while 9 August demonstrated a doubling of the VWS in the northwest quadrant to more than 30 m s−1. It is apparent that the STJ intruded into the 3° latitude ring along the west–north portions while the easternmost extent was sheltered and remained at less than 15 m s−1 VWS. There was less than 5 m s−1 VWS in the southwest quadrant on 9 August, a location beyond the influence of the STJ.
The intrusion of the STJ altered the net divergence encompassed by the circumnavigations. Weak divergence in the upper troposphere on 8 August was replaced by strong convergence that extended downward to nearly 600 hPa by 9 August (Fig. 7).
Based upon examination of satellite image loops, the lower-level circulation was hidden under the high clouds at 0200 UTC 9 August (Fig. 8a). Separation or tilting of the centers, based on the patterns of the lower and upper clouds, was starting to be apparent by 0600 UTC 9 August (Fig. 8b). The lower circulation center, inferred from the cloud patterns, and the upper-level clouds were clearly separated by 1000 UTC (Fig. 8c) and that trend continued (1200 UTC 9 August; Fig. 8d).
Since both satellite imagery and GPS sonde analyses support the contention that the STJ had already intruded into Felicia’s core before the end of the G-IV surveillance on 9 August, the 250-hPa radial inflow along the northwest quadrant was linearly interpolated between the two mission periods, resulting in a value from −10 to −12 m s−1 radial inflow at 0000 UTC 9 August. With the radius of maximum wind (RMW) evaluated at ~80 km, −11 m s−1 radial inflow in the northwest quadrant would have reached Felicia’s RMW in approximately 6.5 h. This RMW of 80 km is based upon the NOAA/Hurricane Research Division’s 9 August H*WIND analysis, which incorporates the C-130 aircraft data and all other available observations (Powell et al. 1998). The timing of this estimated intrusion (~0630 UTC 9 August), corresponding with estimates of cold cloud-top area from IR satellite imagery, is shown in Fig. 9. In Fig. 9, deep convection associated with Felicia is represented by the spatial extent of IR-estimated cloud tops with temperatures less than −50°C collocated within 250 km of the TC center (see Fig. 8). Deep convection covered an area of 3°–4° latitude square until 0600–0700 UTC 9 August, at which point the area <−50°C steadily drops to zero. The arrival time for the STJ at the eyewall is very close in time (±2 h) to the rapid decrease of the area of the cold tops over Felicia’s core.
c. Midlevel dry air is present but remains far from the eyewall
The soundings along the circumnavigations revealed that dry air (low or cool θe) could be found occasionally in the layer from 750 to 500 hPa, with the driest often being near 700 hPa. Storm-relative radial flow in unison with θe at 700 hPa is therefore analyzed with the primary objective to determine if dry air could have been ingested into the eyewall and be responsible for the rapid decay of deep convection there. The storm-relative radial flow demonstrates similar profiles for both 8 and 9 August; peak radial outflow of ~5 m s−1 in the west-northwest sector followed by a gradual decline to very weak radial inflow from −1 to −2 m s−1 in the northeast–southwest portions of the circumnavigation (Fig. 10a). Clockwise from southwest to northeast, both 8 and 9 August indicated that θe fluctuated between 335 and 342 K (Fig. 10b). The southeast quadrant showed a marked decrease to 325–330 K during both sampling periods. This sector of dry air was collocated with very weak storm-relative radial inflow from −1 to −2 m s−1. With the assumption that this radial inflow applies to the majority of the midlevel layer and that the radial inflow does not increase with decreasing radius, this midlevel quadrant of lower θe would have reached Felicia’s RMW in approximately 2 days, a value far too large to be a factor in Felicia’s decay when compared to the collapse of the deep clouds.
d. SHIPS shear evaluation of Felicia
SHIPS is a statistical–dynamical model that employs multiple regression techniques to forecast the intensity of TCs (DeMaria et al. 2005). SHIPS forecasts incorporate data from persistence, climatology, and the environment. Atmospheric predictors such as VWS are retrieved from the NCEP Global Forecast System (GFS) and GOES infrared imagery. SHIPS-forecasted VWS, also known as the SHIPS SHDC variable, is calculated by first removing the 850-hPa GFS vortex center, after which the 850–200-hPa shear magnitude is averaged from 0 to 500 km relative to the 850-hPa vortex center. Since the G-IV datasets on 8 and 9 August were assimilated into the SHIPS 1200 UTC GFS model runs, these prognostications have the added benefit of in situ measurements in the vicinity of Felicia in order to better represent the true environmental flow and, subsequently, VWS.
Figure 11 presents the SHIPS-forecasted shear after maximum intensity at 0600 UTC 6 August. Also plotted are the averaged 850–225-hPa VWS values from the G-IV circumnavigations on 8 and 9 August. While the SHIPS VWS predictors determine a shear value for an area out to a 500-km radius relative to the 850-hPa vortex center, the storm-relative G-IV computations of VWS are solely the average of the values along the 3° latitude rings. Since the values of VWS are not computed in the same fashion, there are large differences between the two approaches. Additionally, inaccuracies in the GFS fields would have a tendency to create more spread between the G-IV-measured VWS and those evaluated from the GFS model. Somewhat surprisingly, the SHIPS forecast on 6 August demonstrated the closest trend with regard to the timing of increasing VWS, though it was late by about 24 h; SHIPS VWS forecasts on 7 and 8 August suggested a reduction in VWS around Felicia from 1200 UTC 8 August to 0000 UTC 9 August while both the SHIPS 6 August forecast and the trend in G-IV circumnavigations from 8 to 9 August portrayed an increasing trend. We feel that the SHIPS calculation of VWS is lower most likely because it calculates the value symmetrically around the TC and includes regions to the south, away from the STJ where there is far less VWS.
The NOAA G-IV circumnavigations have provided insight into the final collapse of Hurricane Felicia (2009). The relative radial flow at 250 hPa increased from −5 to more than −20 m s−1 along the northwest quadrant of the circumnavigation as the range between the TC center and the axis of the subtropical jet stream decreased. The STJ enhanced the vertical shear of the horizontal wind to over 30 m s−1 along the northwest portion of the circumnavigation; the estimated arrival time of this extreme VWS in the eyewall region is well correlated with the dissipation of deep convection located there and new cells developing far downshear. Thereafter, only shallow clouds surrounded the vortex center. Midlevel dry air was present, as could be expected from the hurricane’s location south of the Pacific subtropical high, but this air did not reach the eyewall region during the rapid dissipation and therefore is not deemed a factor in intensity change. Earlier weakening of the TC can be attributed largely to its passage over cooler water.
Based on this study, we recommend two points. First, VWS should be calculated for at least each quadrant of the TC if at all possible; large annuli or full circular area means may mask the invasion of strong winds that result in detrimental VWS. This suggestion follows on with other changes in VWS estimation proposed by Velden and Sears (2014). Second, noting the difference between the actual intensity and the maximum potential intensity is viewed as a useful technique for identifying the influences that contribute to TC weakening beyond theoretical thermodynamic considerations.
The physical mechanism of how exactly VWS disrupts the TC is still unknown. Ventilation and subsequent cooling of the upper-tropospheric warm core (e.g., Simpson and Riehl 1958; Gray 1968) or tilting of the entire warm core column (e.g., DeMaria 1996; Frank and Ritchie 2001) has been hypothesized to be the cause. Neither argument has had the benefit of confirming observations. To better grasp the physics at play, observations in the environment and the warm core, garnered by a Global Hawk flying in the lower stratosphere or by judicious use of the now-Doppler-equipped G-IV, would be highly beneficial. The mission should include simultaneous sampling of the eyewall and lower portions of the vortex by one of the NOAA WP-3D aircraft.
This research is supported by NSF Award AGS-1042680. We appreciate the dedication of the crews of the G-IV of the NOAA-Aircraft Operations Center and the opportunity to fly with them. The authors thank the reviewers for their ideas that improved the manuscript.