The global positioning system (GPS) dropwindsonde (Hock and Franklin 1999) takes high-resolution pressure, temperature, humidity, and wind velocity measurements as it descends from aircraft. It is the culmination of improvements to instruments, tracking techniques, and data transmission that resulted in the balloon-launched radiosonde (or rawinsonde) used today (Fig. 1). This technology had historically been used to obtain vertical soundings upward from the Earth’s surface; in 1949, the U.S. military noted that they could measure surface pressure by releasing a radiosonde without the balloon from aircraft. The instrument was dubbed the dropsonde as the wind-finding capability had not yet been developed. Aircraft crews soon started regular dropsonde releases into tropical cyclone (TC) eyes to measure surface pressure; the first such release in the Atlantic was in Hurricane Able in 1950 (Hagen et al. 2012). A history of the development of the dropsonde is available in Dabberdt et al. (2022).
Timeline of the development of atmospheric sounding technology through the advent of dropwindsondes.
Citation: Bulletin of the American Meteorological Society 104, 11; 10.1175/BAMS-D-22-0119.1
The National Center for Atmospheric Research (NCAR) led the development of a dropwindsonde (renamed due to its ability to measure wind velocity) using the Omega navigation system that became available in 1971 (Govind 1975; Simpson et al. 1975). This Omega dropwindsonde (ODW) was used during the Global Atlantic Research Program (GARP) Atlantic Tropical Experiment (GATE). The real-time data were coarse, mainly at mandatory pressure levels below the aircraft with the later addition of significant levels, and the wind finding (the process by which winds are calculated from the instrument motion) did not consistently work in clouds, prohibiting observations in TC eyewalls, nor during aircraft maneuvers (Franklin and Julian 1985). Postflight data were available at 10-s intervals for thermodynamic data and at 30-s intervals for wind. The ODW did provide the first reliable and frequent soundings in the oceanic TC environment. Joanne Simpson sought other uses of the instrument after GATE. Bob Burpee noted that an important reason for TC forecast failures was the lack of observations in the TC environment. They came up with the idea to drop ODWs from the aircraft for assimilation into operational numerical models, and the ODW was adopted for use by both NOAA and the U.S. Air Force. On 15 September 1982, the first “synoptic flow” mission was conducted around Hurricane Debby (Burpee et al. 1984). Even though only 17 such missions were flown by 1996, improved modeling and data assimilation techniques allowed scientists to show that the ODW data allowed for a 16%–30% reduction in track forecast errors in the cases flown, improvements that were as large as the cumulative error reduction during the previous 20–25 years (Franklin and DeMaria 1992; Burpee et al. 1996). Partly as a result, Congress authorized NOAA to purchase a Gulfstream-IV (G-IV) high-altitude jet to fly operational missions when TCs threatened the United States (Aberson and Franklin 1999). The new GPS technology replaced Omega, and by late 1996, the new GPS dropwindsonde and the Airborne Vertical Atmospheric Profiling System (AVAPS), also developed by NCAR, replaced the ODW (Hock and Franklin 1999).1 The first operational surveillance mission with the new dropwindsonde system was conducted in July 1997 around Tropical Storm Claudette, but the G-IV flew only three missions in the Atlantic that year due to the very inactive season (Aberson and Franklin 1999; Aberson 2002).
On 3 August 1997, during a P-3 research mission in Hurricane Guillermo, Michael Black decided to deploy a GPS dropwindsonde in an eyewall for the first time. The first-ever high-resolution wind and thermodynamic data were obtained in the eyewall, including a wind speed of >78 m s−1 (Lawrence 1999). Since then, the dropwindsonde system has been installed on additional aircraft around the world, including a modified system for the NASA Global Hawk (Wick et al. 2018). Tens of thousands of dropwindsondes have been released in and around TCs worldwide (Fig. 2), revolutionizing TC science, and about 400 research papers have been published in peer-reviewed journals. A continuously updated dropwindsonde data archive in and around TCs is maintained by the Atlantic Oceanographic and Meteorological Laboratory (AOML) Hurricane Research Division (HRD). The data are organized by year, storm name, and aircraft mission and are available at https://www.aoml.noaa.gov/data-products/#hurricanedata, with information on the data formats at https://www.aoml.noaa.gov/hrd/Storm_pages/sondeformat.html. This paper reviews major changes to GPS dropwindsonde use and technology since 1997, and how the data have improved forecasts and changed understanding of TCs.
Storm-relative locations of dropwindsonde releases in various TC stages from 1996 through 2021. (right) A composite of all the dropwindsonde releases. The total number in each life cycle stage is indicated above each panel, with the total number within 500 km of the TC center in parentheses.
Citation: Bulletin of the American Meteorological Society 104, 11; 10.1175/BAMS-D-22-0119.1
History of the GPS dropwindsonde and quality control procedures
The first GPS dropwindsonde (Hock and Franklin 1999), the RD93, was produced from 1996 to 2009. This version used a “codeless” GPS receiver (Kaisti 1995) to calculate the wind velocity. The GPS receiver tracked the relative Doppler frequencies from satellite signals, so no positions or velocities were calculated on the instrument itself. The relative Doppler frequencies corresponded to the combined motions of the satellites and the receiver, and the GPS processing card in the aircraft data system converted them to wind velocity. Starting with the RD94 dropwindsonde, a GPS receiver that calculates the location of the instrument itself has been used, thus allowing for more accurate and reliable wind velocity measurements than before, especially in high wind speed regions near the surface. That version also allowed for an increased data rate for the wind velocity. The subsequent upgrade to the RD41 system in 2018 provided additional improvements. Miniaturized versions of the last two (NRD94 and NRD41) have also been manufactured, and the NRD41 dropwindsonde is due to become the main instrument in TC operations in 2023. Some instruments were outfitted with an infrared (IR) sensor (Fig. 3) to provide sea surface2 temperature measurements along with the other observations. Specifications of these instruments are provided in Table 1.
(left) View of a standard dropwindsonde and (right) inside view of all sensors showing location of additional IR sensor. From Zhang et al. (2017).
Citation: Bulletin of the American Meteorological Society 104, 11; 10.1175/BAMS-D-22-0119.1
Specifications of the various GPS dropwindsonde systems in use. Repeatability refers to the standard deviation of differences between two successive repeated calibrations. (Information courtesy of Vaisala, Inc., manufacturer of the RD instruments, and NCAR, producer of the NRD instruments.)
The original processing and quality-control software, known as “editsonde,” was written and maintained by AOML/HRD and installed on the NOAA aircraft to facilitate real-time data transmission so that the data could be used by forecasters and be assimilated into numerical models. Editsonde was also used postflight to create research-quality datasets. Because the graphical interface was designed to use Hewlett-Packard libraries (Griffin et al. 1992), editsonde could not be converted easily to other computer systems. NCAR developed the Atmospheric Sounding Processing Environment (ASPEN), based on editsonde algorithms, which is now used on all aircraft and for postflight research. Editsonde required frequent manual intervention during quality control; ASPEN was designed to operate as automatically as possible using a graphical user interface, while allowing for manual intervention when required.
Both software packages identify the sounding termination (usually splash) point and apply an automated algorithm sequence to flag erroneous data. Corrections for the wind shear and thermal-sensor lags are applied, and a filter eliminates aliasing of unresolvable scales (i.e., scales of ∼3 s or less) (Hock and Franklin 1999). The final products are the quality-controlled data in multiple formats including TEMP DROP, Binary Universal Form for the Representation of meteorological data (BUFR), full-resolution data (FRD) files, and Network Common Data Form (NetCDF) files. Skew T–logp diagrams of individual soundings and synoptic station plots can be produced, and other files contain information on the quality-control procedures that were used.
Real-time and best-track intensity estimation
Before the advent of the GPS dropwindsonde, the main technique to estimate TC intensity (defined as the maximum sustained, or time-averaged, wind speed at 10-m altitude) from aircraft measurements was to adjust aircraft flight-level wind speed measurements downward by 10%–20% (Franklin et al. 2003). The first operational use of dropwindsonde data in the TC eyewall occurred in Hurricane Georges (1998) when it was near peak intensity. NHC’s Tropical Cyclone Discussions made numerous references to eyewall dropsonde wind reports, and in postanalysis, Georges’ assigned intensity of 135 kt (1 kt ≈ 0.51 m s−1) was based largely on these data. The measured low-level wind speed was comparable to that at the 3-km flight level, but subsequent observations in Georges during its weakening phase reported low-level wind speed 20%–30% below that at flight level (Pasch et al. 2001). Franklin et al. (2003) quantified adjustments between maximum flight-level and surface wind speeds based on dropwindsonde data and noted that they varied with azimuth and distance from the TC center. Noting that the dropwindsonde measures effectively instantaneous wind velocity, whereas operational centers quantify intensity based on a sustained (time-averaged) wind speed, they calculated a relationship between the mean wind speed in the lowest 150 m of the sounding (corresponding to measurements averaged for about 15 s), the height of the midpoint of that layer, and the wind speed at the surface; Fig. 4 shows the current relationship used at the National Hurricane Center for poststorm analysis, and is a close approximation to what is used in operations. The mean wind speed calculated in the lowest 150 m of the profile was subsequently included in the TEMP-DROP message, and these data have been instrumental in real-time analysis of TC intensity. Even with the advent of the Stepped-Frequency Microwave Radiometer (SFMR; Uhlhorn and Black 2003), such wind speed reductions remain important for TC intensity diagnosis.
Reduction factor between the mean wind speed in the lowest 150 m of a dropwindsonde sounding and the surface wind speed in relation to the height of the midpoint of the 150-m layer. This height varies due to failures of the instrument before reaching the surface. This factor is used operationally at the National and Central Pacific Hurricane Centers.
Citation: Bulletin of the American Meteorological Society 104, 11; 10.1175/BAMS-D-22-0119.1
The varying and uncertain reduction factor from flight level to the surface is a major factor in the Atlantic hurricane reanalysis project, a reevaluation of past TC tracks and intensities like that of Hurricane Andrew (Dunion et al. 2003; Landsea et al. 2004, 2008, 2012, 2014; Kieper et al. 2016; Hagen et al. 2012; Delgado et al. 2018). Further research into reductions of flight-level wind measurements to the surface (e.g., Powell et al. 2009; Uhlhorn et al. 2014) and refinements to surface wind speed retrievals from SFMR are ongoing.
Observing system validation
Because they are in situ measurements with known error characteristics, dropwindsonde data are used for comparison with, and calibration of, other observing systems critical for real-time and postseason TC intensity and structure analyses and for assimilation into numerical forecast. They have validated atmospheric motion vectors (Dunion and Velden 2002; Dunion et al. 2002; Sears and Velden 2012), surface wind speed measurements from the SFMR (Uhlhorn and Black 2003; Uhlhorn et al. 2007; Powell et al. 2009; Klotz and Uhlhorn 2014; Holbach et al. 2018; Sapp et al. 2019), Aqua Moderate Resolution Imaging Spectroradiometer profiles (Zhang et al. 2006), Doppler wind lidar measurements (Bucci et al. 2018; Zhang et al. 2018), Special Sensor Microwave Imager and Quick Scatterometer surface wind speed observations (Chen 2007; Chou et al. 2010), and Synthetic Aperture Radar retrieved wind directions (Gao et al. 2018). They validate high wind speed model functions for scatterometers (Brown and Zeng 2002) and other dropwindsonde systems such as the High-Definition Sounding System (Black et al. 2017). Cione et al. (2016, 2020) used dropwindsondes to corroborate observations from small uncrewed aircraft deployed in Hurricanes Edouard (2014), Maria (2017), and Michael (2018). Dropwindsonde data composites confirmed other near-surface boundary layer thermodynamic measurements (Wadler et al. 2022).
Forecast impact, data assimilation, and targeted observations
NOAA models with data assimilation systems incorporated the dropwindsonde temperature and wind information as soon as the data became available in 1997; moisture data were not used due to potential observational error and deficiencies in the assimilation systems. During the first few years of operational synoptic surveillance with the G-IV, only modest improvements to model forecasts of TC track and intensity were seen. The issue was not the dropwindsonde data themselves, but in how they were assimilated into the models with the then-current three-dimensional variational scheme (Aberson 2002). Aberson (2008) noted forecast degradations due to the assimilation of erroneous data and data in the TC eyewall (that the models at the time could not represent due to their low resolutions) into the models. Operational data processing, quality control, and data assimilation systems have been improved to address these problems (Aberson 2010).
Observing strategies to optimize and maximize the impact of dropwindsonde observations on numerical model forecasts have been developed. In Aberson (2003), the fastest growing modes, or targets, were identified by areas of large forecast spread in the NCEP ensemble forecasting system; the observing strategy required regularly spaced sampling of the entire target due to the 3D-VAR scheme with isotropic covariances. This targeting technique provided larger improvements with the assimilation of the targeted data subset than when the entire dropwindsonde datasets were assimilated, highlighting a weakness of that data assimilation scheme. That technique, along with a companion scheme utilizing a solution to the traveling-salesman problem (Press et al. 2007), was implemented into operations during the 2008 Atlantic hurricane season, allowing for optimal coverage of the target. Other targeting techniques (Aberson 2010; Reynolds et al. 2013; Majumdar et al. 2013; Wu et al. 2014; Torn 2014) have been developed. Most recently, Ditchek and Sippel (2023) in an observing system experiment found that dropwindsonde data within the TC vortex were most important to intensity and structure forecasts and that dropwindsonde data should be obtained close to the center (i.e., the eye and eyewall) of intense TCs for optimal improvement. They also found that not sampling a TC at all is better than only sampling the inner core, confirming the 2018 decision to regularly release dropwindsondes at the end of each radial leg. Further improvements to the operational targeting technique (e.g., Ancell and Hakim 2007; Torn and Hakim 2008) are being tested for operational implementation.
Additional forecast improvements coincided with improved assimilation techniques and assimilation of additional data and the ability to obtain observations from altitudes near and above the tropopause. The first use of dropwindsonde humidity data in global models led to forecast track improvements as large as those from assimilating the wind data (Kamineni et al. 2006); NCEP began assimilating dropwindsonde humidity measurements into their operational models a few years later. Since dropwindsondes are advected by the wind, they have been seen to drift more than halfway around the eyewall in intense TCs; only the launch location is available in the original TEMP DROP message, so a technique to calculate the location of each observation in the message (Aberson et al. 2017a) allows for the assimilation of dropwindsonde data in high-resolution model systems. Christophersen et al. (2017, 2018a,b) and Kren et al. (2018) assessed the impact of dropwindsonde data in the stratosphere and upper troposphere during the Hurricane and Severe Storm Sentinel (Braun et al. 2016) and Sensing Hazards with Operational Unmanned Technology (Wick et al. 2018) programs, confirming the additive impact of dropwindsonde data when mixed with other data types like those now available from low-orbit satellites. The first attempt to assimilate dropwindsonde data using an ensemble Kalman filter (Aberson and Etherton 2006) was soon followed by significant improvements to that methodology (Torn and Hakim 2008) and implementation of ensemble-based data assimilation systems in numerical model systems globally.
The success of targeted dropwindsonde observations in the Atlantic and eastern North Pacific basins led Taiwan to institute their own aircraft-based program, the Dropwindsonde Observations for Typhoon Surveillance near the Taiwan Region (DOTSTAR). Other programs in the western North Pacific basin followed (Wu et al. 2007a,b; Chou and Wu 2008; Yamaguchi et al. 2009; Reynolds et al. 2010; Chou et al. 2011; Xie et al. 2013; Holt et al. 2015).
Characterizing and understanding physical structures and processes
Dropwindsonde data have greatly improved understanding of TC dynamics and thermodynamics, leading to improvements to models and parameterization schemes and official forecasts of track, intensity, and structure. Some understandings may not translate directly to changes to operational models or products, but may be incorporated indirectly into manual forecast techniques used by specialists to make forecasts, especially of intensity and structure. The data have been used directly to validate and improve models that are the bases for official forecasts.
High-resolution observations in the TC boundary layer were very rare until the late 1990s. Wind speed within the boundary layer was thought to decrease from gradient flow at the top with a progressive turning of the wind toward the TC center near the surface, though some previous observations suggested the presence of a supergradient low-level jet (Kepert 2001). Dropwindsonde data show substantial variability in boundary layer structure, especially in the inflow angle (Zhang and Uhlhorn 2012), and that a low-level wind speed maximum is common (Kepert 2001; Kepert and Wang 2002; Giammanco et al. 2013). The heights of this jet and of the boundary layer top were found to decrease with decreasing radius from the TC center, and the jet is strongest on the motion-relative left side of the TC (Kepert 2006a,b; Schwendike and Kepert 2008; Zhang et al. 2011). Improved boundary layer models were developed as a result (Vickery et al. 2009; Lee and Chen 2012; Snaiki and Wu 2018) and underscored the threat to high-rise structures noted by Franklin et al. (2003). Emergency managers had once suggested “vertical evacuation” to keep people safe from storm surge, but these new findings changed these recommendations to make such evacuation a last resort.
Knowledge of the mechanisms by which heat, moisture, and momentum are transferred between the atmosphere and the ocean is necessary to improve numerical model forecasts, especially of TC intensity. Since high-resolution near-surface wind measurements were difficult to obtain in high-wind regimes, values of parameters such as surface stress, roughness length, and drag coefficients were extrapolated from measurements in low wind speed regimes. Parameters such as momentum and enthalpy fluxes, roughness lengths, and exchange coefficients were estimated for the first time in intense TCs using dropwindsonde data (Fig. 5) through profile (Powell et al. 2003; Holthuijsen et al. 2012; Richter and Stern 2014; Richter et al. 2016, 2021) and budget (Bell et al. 2012) methods; vertical eddy diffusivity in the boundary layer was derived with flight-level data using a flux calculation method and dropwindsonde data for a vertical gradient calculation method (Zhang et al. 2011; Zhang and Drennan 2012). The various parameters change with wind velocity, TC region, and sea state. Zhang et al. (2015) presented a feedback framework by which aircraft observations in TCs are used to identify model deficiencies, and new schemes are developed based on these data, with subsequent improvements to model systems, especially in the boundary layer. This developmental framework was successful in improving physics related to turbulent transfer processes, surface layer, boundary layer, and horizontal diffusion parameterization schemes in the operational Hurricane Weather Research and Forecast (HWRF) model, resulting in improved track and intensity forecasts (Zhang et al. 2015). Zhang et al. (2015, 2017) studied the impact of boundary layer physics improvement on HWRF forecasts of two TCs undergoing rapid intensification (RI) and found that observation-based modification of the vertical eddy diffusivity (Km) in the planetary boundary layer (PBL) scheme (Fig. 6) substantially improved the RI forecasts. The data have been used for model evaluation, especially for new model systems. Liu et al. (2012) used the data to verify forecast models beyond the regular parameters used in TC model evaluation. Bryan et al. (2017), Stern and Bryan (2018), and Stern et al. (2021) used dropwindsonde data to evaluate large-eddy TC simulations, especially in the boundary layer. Details of how the observations led to improvements to operational TC models are presented in Zhang et al. (2020).
(left) Neutral drag coefficient (Cd10N) as a function of 10-m wind speed (U10). The green circles show results from Bell et al. (2012). Black and blue symbols indicate data adapted from French et al. (2007) and Vickery et al. (2009), respectively. The crosses show mean and standard deviation of Cd10N at a given wind speed. (right) Exchange coefficient for enthalpy transfer as a function of U10 from multiple studies. The left and right panels were taken from Bell et al. (2012) and Richter and Stern (2014), respectively.
Citation: Bulletin of the American Meteorological Society 104, 11; 10.1175/BAMS-D-22-0119.1
Height–distance (normalized by the radius of maximum wind speed) schematics showing different structures in Hurricane Weather Research and Forecast Model forecasts with (left) large vertical eddy diffusivity (Km) and (right) small Km. The green line denotes the boundary layer height (h). The black dashed line shows the radius of maximum wind speed varying with height (z). The black solid lines are contours of the tangential wind speed, showing the vortex depth. This figure is a modified version of Fig. 17 in Zhang et al. (2015).
Citation: Bulletin of the American Meteorological Society 104, 11; 10.1175/BAMS-D-22-0119.1
Dropwindsonde data helped to define different boundary layers (defined by thermodynamics, kinematics, or multiple parameters) in the TC and allowed for the first detailed look at the thermodynamics and energetics in the TC inflow layer. Traditional methods to define the boundary layer based on the Richardson number are not appropriate to TCs (Fig. 7; Zhang et al. 2011). The height of the top of the boundary layer varies by quadrant (Zhang et al. 2009, 2011, 2013; Ming et al. 2015). Dropwindsonde data illuminated changes to the thermodynamics and energetics along inflow trajectories within the boundary layer (Wroe and Barnes 2003; Schneider and Barnes 2005; Barnes 2008; Barnes and Dolling 2013; Fig. 8) supporting hypotheses that hurricane intensity appears to be strongly modulated by energy exchange near and under the eyewall. Equivalent potential temperature was discovered to be unmixed in the boundary layer several hundred km from the TC center; this was contrary to then-current theory but supported similar findings in numerical simulations (Montgomery et al. 2014). Outside the core, cool and dry convective downdrafts reduce enthalpy in the boundary layer as it is gained from the warm sea below. Surface fluxes increase close to the eyewall, but little energy loss through the layer top is seen. It appears that the TC secondary circulation enhances convection in this region, thus allowing for an increase in energy and supporting the hypothesis that TC intensity is controlled by energy exchanges close to and under the eyewall. The radial wind decelerates in the boundary layer leading to low-level convergence just inward of the eyewall. This leads to deep convection in the eyewall and is similar whether it occurs in the primary or secondary eyewall during a replacement cycle (Bell et al. 2012; Abarca et al. 2016). These studies and others (Montgomery et al. 2014; Rogers et al. 2015, 2016; Zawislak et al. 2016) found gradient imbalances in dropwindsonde data during intensification.
Schematic of boundary layer height with distance from the center normalized by the radius of maximum wind speed (RMW) based on dropwindsonde composites. Here, hinfl is the depth of the inflow layer, zi is the depth of the mixed layer, and hυmax is the height of the maximum wind speed. The solid black line denotes the height where the bulk Richardson number is equal to 0.25. From Zhang et al. (2011).
Citation: Bulletin of the American Meteorological Society 104, 11; 10.1175/BAMS-D-22-0119.1
Schematic of an equivalent potential temperature (θe) budget along the inflow trajectory in the boundary layer of Hurricane Bonnie (1998). This figure is from Wroe and Barnes (2003) and is based on dropwindsonde data.
Citation: Bulletin of the American Meteorological Society 104, 11; 10.1175/BAMS-D-22-0119.1
TCs in low-shear environments tend to intensify, and those in highly sheared environments tend to weaken. In moderately shear environments (9–22 kt; Rios-Berrios and Torn 2017), TCs do everything from rapidly intensify to weaken, increasing the difficulty of some forecasts. Rogers et al. (2015, 2016), Zawislak et al. (2016), and Nguyen et al. (2017) found that shear-related asymmetries in precipitation, humidity, entropy, and temperature contribute to TC intensity changes (Fig. 9). Dropwindsonde data showed that dry layers originate from dynamically forced and precipitation-driven subsidence in the upshear hemisphere; these layers hinder intensification as they increase stability, cap the boundary layer, and suppress convection. Nguyen et al. (2017), Wadler et al. (2018), and Rogers et al. (2020) found that convective downdrafts in the left-of-shear and upshear quadrants sometimes carry low-entropy air downward, cooling and stabilizing the low levels downstream. This stabilization can inhibit the development of precipitation symmetrically around the storm, limiting intensification. However, the air recovers entropy through surface enthalpy fluxes as it circulates toward the downshear-right quadrant, favoring intensification (Zhang et al. 2017; Nguyen et al. 2019; Wadler et al. 2022), and midtropospheric moistening from convection may remove dry layers aloft (Zawislak et al. 2016). Nguyen et al. (2019) found, by compositing a large number of dropwindsonde observations in and around weak TCs, that conditional instability is maximized downshear and minimized upshear regardless of intensity change, but that stability decreases with intensification rate due to increased moist entropy in the boundary layer.
Three-dimensional schematics summarizing hypothesized hindrances to precipitation symmetry in TCs in vertical wind shear (denoted by the black arrow in each panel). (top) An asymmetric case (Cristobal) and (bottom) a symmetric case (Bertha). The quadrants are denoted downshear right (DSR), downshear left (DSL), upshear left (USL), and upshear right (USR). Cloud features show locations of clouds and precipitation, and shading denotes regions of high (light red) and low (light blue) boundary layer θe. Dashed arrows denote updrafts (red) and downdrafts (blue). Green arrows show surface enthalpy fluxes; green circles in the bottom panel denote detrainment from clouds. Dark orange shading denotes lateral advection; light orange shading denotes mesoscale subsidence. Adapted from Nguyen et al. (2017).
Citation: Bulletin of the American Meteorological Society 104, 11; 10.1175/BAMS-D-22-0119.1
Dropwindsondes have observed mesoscale and misoscale features in eyes and eyewalls of intense TCs, though it remains unclear whether these features have any impact on TC intensity or structure. These features may increase TC intensity and produce vertical motions that create hazards for reconnaissance aircraft and may be responsible for catastrophic damage over land (Montgomery et al. 2006; Aberson et al. 2006; Bell and Montgomery 2008; Sanger et al. 2014; Nugent and Rios-Berrios 2018). Stern et al. (2016) found that these features, characterized by high horizontal and vertical wind velocities, are likely ubiquitous in eyewalls of very intense TCs and that the strongest updrafts measured by dropwindsondes occur mainly to the left of the shear vector and in very intense, steady-state TCs; few occur in TCs that are undergoing rapid intensification. The strongest updraft and downdraft measured by dropwindsonde (27.4 and −30.1 m s−1) were in Hurricanes Patricia and Felix at the time their rapid intensification had ended (Rogers et al. 2017; Aberson et al. 2017b). Moist entropy differences between the eye and eyewall show that advection from the eye to the eyewall can trigger intensification, especially in weak systems (Barnes and Fuentes 2010; Dolling and Barnes 2012), but the fact that the eye entropy is quickly depleted suggests that this process is rare. Strong eyewall downdrafts are not made up of low-entropy air, possibly because they are associated with a large amount of evaporation; in contrast, weak downdrafts reaching the boundary layer are made of low-entropy air, suggesting that they are a greater hindrance to intensification than the strongest downdrafts (Wadler et al. 2018). TCs have some of the largest recorded helicity values, suggesting that supercells exist in TCs (Molinari and Vollaro 2008, 2010; Molinari et al. 2012). Baker et al. (2009) and Eastin and Link (2009) looked at whether midlatitude paradigms for supercells and tornadoes can be applied to landfalling TCs and found dry-air intrusions that affect the instability of mesoscale convective systems in TCs. TC intensity changes are related to the vertical motion in the eyewall (Lowag et al. 2008). Eastin et al. (2005a,b) documented the contribution of buoyant updrafts to the total upward mass transport using dropwindsonde and flight-level data. The low-level eye is an important source for buoyant updrafts in the eyewall via the export of high-entropy air, in addition to the low-level inflow from outside the TC. Most buoyant updrafts are located downshear left of the TC center, whereas convective downdraft cores are found upshear (Eastin et al. 2005a,b). Stern et al. (2016) found that the strongest low-level updrafts are not caused by buoyancy but agreed that they mainly occur to the left of the shear vector and in very intense, steady-state TCs.
Dropwindsonde observations led to great strides in identifying and understanding thermodynamic structures in both the large-scale tropical environment and within TCs themselves. Dunion and Velden (2004) used dropwindsonde data to show that the Jordan mean tropical sounding, said to represent the climatology in the tropical North Atlantic and Caribbean Sea, did not well represent the complex moisture distribution in the region, and new mean soundings for use in idealized models and other research were developed. Instead, a moist environment periodically modified by dry-air outbreaks from the Sahara and midlatitudes is evident. Forecast models were found to be too moist where there are Saharan air layers, and this bias can have a large impact on TC intensity and genesis forecasts (Zipser et al. 2009; Folmer et al. 2016).
Data from dropwindsondes released by high-altitude aircraft provided key insights into the structures of the upper-level outflow layer. Molinari et al. (2014) and Molinari and Vollaro (2014) found three turbulent layers associated with the upper-level outflow near the tropopause. The central dense overcast creates its own stability profile that influences the turbulence and the transition to outflow. Other upper-level sampling in a weak, sheared TC during the Fourth Convection and Moisture Experiment (CAMEX-4), especially within the eye, validated structures seen in mesoscale models (Halverson et al. 2006). The low surface pressure was accounted for by vertical layers of warming within the eye, and asymmetry was related to the interaction of the vortex with vertical wind shear. The observations confirmed model simulations and theory of sheared TCs (Heymsfield et al. 2006). Later high-altitude dropwindsonde data collected in six TCs during NASA’s Hurricane and Severe Storm Sentinel experiment provided high-resolution observations of TC outflow and the warm core. The amplitude, but not the height, of the warm core was correlated with TC intensity. The outflow potential temperature was correlated with that of the low-level inflow rather than with TC intensity (Komaromi and Doyle 2017).
Dropwindsonde data have been critically important to the study of tropical cyclogenesis. Systems undergoing cyclogenesis have been sampled by aircraft far less frequently than developed TCs (Zawislak et al. 2022). Much of our current knowledge is from intensive airborne observing programs in 2006 [NASA’s extension to the African Monsoon Multidisciplinary Analysis experiment (NAMMA; Zipser et al. 2009)] and 2010 [NASA’s Genesis and Rapid Intensification Processes (GRIP; Braun et al. 2013) and the Pre-depression Investigation of Cloud Systems in the tropics (PREDICT) experiments; Montgomery et al. 2012]. Dropwindsonde data from NAMMA characterized waves as they emerged from the African coast and showed that those that develop into TCs exhibit small-amplitude midlevel troughs that were unlikely to ingest dry, dusty Saharan air (Zawislak and Zipser 2010). Dropwindsonde data from PREDICT and GRIP showed that developing disturbances were moister than nondeveloping ones and that moisture and stability both increased with time; nondeveloping disturbances dried and become more unstable in time (Zawislak and Zipser 2014, Fig. 10). Vertical misalignment between low- and midlevel vortices made systems susceptible to dry-air entrainment (Davis and Ahijevych 2012, 2013; Komaromi 2013; Penny et al. 2015; Helms and Hart 2015; Brammer et al. 2018). Developing disturbances had large positive temperature anomalies that grew vertically and deepened, sometimes starting in the low to middle troposphere (Kerns and Chen 2015), and a strong upper-level radial temperature gradient. TC formation followed a path by which a midlevel vortex in a moist, humid environment precedes low-level vortex amplification (Davis et al. 2014; Zawislak and Zipser 2014). This pathway evolved such that the troposphere stabilizes (low-level cooling and upper-level warming), allowing the peak in the vertical mass flux to move to low altitudes (Raymond et al. 2011; Raymond and López Carrillo 2011). This process enhanced low-level mass convergence and vorticity, spinning up the vortex and suppressing lateral moisture export.
Mean profiles of θe, water vapor mixing ratio (WVMR), relative humidity (RH), and potential temperature θ for nondeveloping (NDEV; solid), developing pregenesis (DEV B; dashed), and developing postgenesis (DEV A; dotted) cases. Profiles are separated for the core (within 333 km of the center; black) and environment (333–778 km of the center; gray). The approximate number of samples in each profile is in parentheses. From Zawislak and Zipser (2014).
Citation: Bulletin of the American Meteorological Society 104, 11; 10.1175/BAMS-D-22-0119.1
Dropwindsonde data have been used to study the end of the TC life cycle when they decay into remnant lows or transition into extratropical cyclones. Canada’s Convair 580 aircraft conducted the first sampling of a TC undergoing extratropical transition in 2000. Abraham et al. (2004) and Fogarty et al. (2006) found a low-level stable layer over the cool ocean, large vertical gradients in horizontal wind speed, and a large area of strong low-level wind on the eastern side of the TC. They further used the data to validate numerical simulations. Dropwindsonde sampling and study of TCs undergoing extratropical transition is ongoing.
The future
The impacts of GPS dropwindsonde data on TC climatology, forecasts, and understanding should not be underestimated. The dropwindsonde has provided researchers and operational scientists with accurate, high-resolution kinematic and thermodynamic observations since 1996. Observations in the turbulent boundary layer that were previously difficult to obtain have been especially important in improving physical parameterizations and numerical guidance. Improved physical understanding, advanced parameterization schemes, and enhanced data assimilation have resulted in improved forecasts and warnings. These improvements are summarized in this review.
The question now is, what’s next? One change in the immediate future is to the instrument itself and the data that are transmitted. The NRD41 dropwindsonde will become the standard dropwindsonde during the 2023 hurricane season, and automatic launch mechanisms for this instrument are under development. Upgrades to aircraft-to-ground communications will allow for the transmission of the full high-resolution data in BUFR format to the ground allowing specialists to see small-scale features measured by the instruments and for the most data to be assimilated into model systems. And implementation of the IR sensor on all instruments to regularly measure sea surface temperature is possible.
Operational and research use of GPS dropwindsondes in TCs will continue for the foreseeable future, as they make up a vital part of the current mix of observations. One limitation of dropwindsondes is that they remain aloft for a limited time period with residence times for the instrument of about 5 min from the standard flight altitude inside the TC (∼8–10 kft) and up to 20 min from high-altitude aircraft (>40 kft); the residence time is only about 1 min within the boundary layer. Innovations using AVAPS can help to solve this problem. Small Uncrewed Aerial Systems (sUASs) may allow for anywhere between 1.5 and 4 h of residence time within the boundary layer, increasing the total number of observations and expanding coverage within the TC. The sUASs will be able to measure sea surface temperature using an onboard IR sensor, wave height/spectra via altimetry, and high-frequency wind velocity using multihole probes, allowing for direct momentum-flux measurements using an eddy correlation method. It is worth noting that the sUAS’s IR/SST capability can be incorporated onto the existing dropwindsonde platform for minimal cost and engineering outlay. Coyote sUAS flights in major Hurricanes Edouard, Maria, and Michael (Cione et al. 2020) and an Altius 6003 flight into major Hurricane Ian suggest that the future is bright for this emerging technology.
Other potential technologies that could increase the volumetric and temporal coverage of observations within TC environments are Skyfora’s StreamSonde and WindBorne balloons. StreamSonde is a promising air-deployed observing system that could provide valuable data using a swarm of very light, small sensors that stay aloft and transmit data for 30–60 min or longer. This will be tested in clear air by NOAA in 2023 and remains as an intriguing possibility to explore in TC environments in the future. WindBorne, a potentially global constellation of low-cost, long-duration (on the order of 1–2 weeks), weather balloons that can be deployed from land, sea, and air, can provide observations throughout the troposphere. This system was tested using land-based deployments during the 2022 Atlantic Hurricane season. A major focus of the Advancing the Prediction of Hurricanes Experiment (APHEX; Zawislak et al. 2022) is to explore new observing systems, and other encouraging technologies are in development. NOAA and its partner organizations will continue to investigate and invest in the most promising examples of these new emerging systems as they become available.
An instrument based on the Long Range Navigation (LORAN) system was also developed, but not used in TC research or operations.
The instrument may also provide the temperature of cloud tops if it is released in clear air, though this has not been tested extensively.
Information on the Altius 600 can be found at https://areai.com/altius-600-2/.
Acknowledgments.
The authors thank Trey Alvey, Frank Marks, James Franklin, Holger Vömel, and an anonymous reviewer for helpful comments on various versions of the manuscript. Mike Jankulak helped edit the manuscript. Jun Zhang acknowledges support from NOAA Grants NA21OAR4590370, NA22OAR4590118, and NA22OAR4590178; NSF Grant AGS2211308; and ONR Grant N00014-20-1-2071. Jonathan Zawislak acknowledges support from NASA Grant 80NSSC19K0012 under the Weather and Atmospheric Dynamics program and TCRI ONR BAA N00014-19-S-B001. Kathryn Sellwood acknowledges funding from TCRI ONR BAA N00014-19-S-B001. The scientific results and conclusions, as well as any views or opinions expressed herein, are those of the author(s) and do not necessarily reflect those of OAR or the Department of Commerce.
Data availability statement.
No datasets were generated or analyzed during the current study. A continuously updated dropwindsonde data archive in and around TCs is maintained by AOML/HRD. The data are organized by year, storm name, and aircraft mission and are available at https://www.aoml.noaa.gov/data-products/#hurricanedata, with information on the data formats at https://www.aoml.noaa.gov/hrd/Storm_pages/sondeformat.html.
References
Abarca, S. F., M. T. Montgomery, S. A. Braun, and J. Dunion, 2016: On the secondary eyewall formation of Hurricane Edouard (2014). Mon. Wea. Rev., 144, 3321–3331, https://doi.org/10.1175/MWR-D-15-0421.1.
Aberson, S. D., 2002: Two years of operational hurricane synoptic surveillance. Wea. Forecasting, 17, 1101–1110, https://doi.org/10.1175/1520-0434(2002)017<1101:TYOOHS>2.0.CO;2.
Aberson, S. D., 2003: Targeted observations to improve operational tropical cyclone track forecast guidance. Mon. Wea. Rev., 131, 1613–1628, https://doi.org/10.1175//2550.1.
Aberson, S. D., 2008: Large forecast degradations due to synoptic surveillance during the 2004 and 2005 hurricane seasons. Mon. Wea. Rev., 136, 3138–3150, https://doi.org/10.1175/2007MWR2192.1.
Aberson, S. D., 2010: 10 years of hurricane synoptic surveillance (1997–2006). Mon. Wea. Rev., 138, 1536–1549, https://doi.org/10.1175/2009MWR3090.1.
Aberson, S. D., and J. L. Franklin, 1999: Impact on hurricane track and intensity forecasts of GPS dropwindsonde observations from the first-season flights of the NOAA Gulfstream-IV jet aircraft. Bull. Amer. Meteor. Soc., 80, 421–428, https://doi.org/10.1175/1520-0477(1999)080<0421:IOHTAI>2.0.CO;2.
Aberson, S. D., and B. J. Etherton, 2006: Targeting and data assimilation studies during Hurricane Humberto (2001). J. Atmos. Sci., 62, 175–186, https://doi.org/10.1175/JAS3594.1.
Aberson, S. D., M. T. Montgomery, M. M. Bell, and M. L. Black, 2006: Hurricane Isabel (2003): New insights into the physics of intense storms. Part II: Extreme localized wind. Bull. Amer. Meteor. Soc., 87, 1349–1354, https://doi.org/10.1175/BAMS-87-10-1349.
Aberson, S. D., K. J. Sellwood, and P. A. Leighton, 2017a: Calculating dropwindsonde location and time from TEMP-DROP messages for accurate assimilation and analysis. J. Atmos. Oceanic Technol., 34, 1673–1678, https://doi.org/10.1175/JTECH-D-17-0023.1.
Aberson, S. D., J. A. Zhang, and K. N. Ocasio, 2017b: An extreme event in the eyewall of Hurricane Felix on 2 September 2007. Mon. Wea. Rev., 145, 2083–2092, https://doi.org/10.1175/MWR-D-16-0364.1.
Abraham, J., J. W. Strapp, C. Fogarty, and M. Wolde, 2004: Extratropical transition of Hurricane Michael: An aircraft investigation. Bull. Amer. Meteor. Soc., 85, 1323–1339, https://doi.org/10.1175/BAMS-85-9-1323.
Ancell, B., and G. J. Hakim, 2007: Comparing adjoint and ensemble sensitivity analysis with applications to observation targeting. Mon. Wea. Rev., 135, 4117–4134, https://doi.org/10.1175/2007MWR1904.1.
Baker, A. K., M. D. Parker, and M. D. Eastin, 2009: Environmental ingredients for supercell and tornadoes within Hurricane Ivan. Wea. Forecasting, 24, 223–244, https://doi.org/10.1175/2008WAF2222146.1.
Barnes, G. M., 2008: Atypical thermodynamic profiles in hurricanes. Mon. Wea. Rev., 136, 631–643, https://doi.org/10.1175/2007MWR2033.1.
Barnes, G. M., and P. Fuentes, 2010: Eye excess energy and the rapid intensification of Hurricane Lili (2002). Mon. Wea. Rev., 138, 1446–1458, https://doi.org/10.1175/2009MWR3145.1.
Barnes, G. M., and K. P. Dolling, 2013: The inflow to tropical Cyclone Humberto (2001) as viewed with azimuth–height surfaces over three days. Mon. Wea. Rev., 141, 1324–1336, https://doi.org/10.1175/MWR-D-11-00348.1.
Bell, M. M., and M. T. Montgomery, 2008: Observed structure, evolution, and potential intensity of category 5 Hurricane Isabel (2003) from 12 to 14 September. Mon. Wea. Rev., 136, 2023–2046, https://doi.org/10.1175/2007MWR1858.1.
Bell, M. M., M. T. Montgomery, and K. A. Emanuel, 2012: Air-sea enthalpy and momentum exchange at major hurricane wind speeds observed during CBLAST. J. Atmos. Sci., 69, 3197–3222, https://doi.org/10.1175/JAS-D-11-0276.1.
Black, P. L., L. Harrison, M. Beaubien, R. Bluth, R. Woods, A. Penny, R. W. Smith, and J. D. Doyle, 2017: High-Definition Sounding System (HDSS) for atmospheric profiling. J. Atmos. Oceanic Technol., 34, 777–796, https://doi.org/10.1175/JTECH-D-14-00210.1.
Brammer, A., C. D. Thorncroft, and J. P. Dunion, 2018: Observations and predictability of a nondeveloping tropical disturbance of the eastern Atlantic. Mon. Wea. Rev., 146, 3079–<