Abstract

Ocean wind vectors from the SeaWinds scatterometer aboard the Quick Scatterometer (QuikSCAT) satellite and Geostationary Operational Environmental Satellite (GOES) imagery are used to develop an objective technique that can detect and monitor tropical disturbances associated with the early stages of tropical cyclogenesis in the Atlantic basin. The technique is based on identification of surface vorticity and wind speed signatures that exceed certain threshold magnitudes, with vorticity averaged over an appropriate spatial scale. The threshold values applied herein are determined from the precursors of 15 tropical cyclones during the 1999–2004 Atlantic Ocean hurricane seasons using research-quality QuikSCAT data. The choice of these thresholds is complicated by the lack of suitable validation data. The combination of GOES and QuikSCAT data is used to track the tropical disturbances that are precursors to the 15 tropical cyclones. This combination of data can be used to test detection but is not as easily used to examine false alarms. Tropical disturbances are found for these cases within a range of 19–101 h before classification as tropical cyclones by the National Hurricane Center. The 15 cases are further subdivided based upon their origination source (i.e., easterly wave, upper-level cutoff low, stagnant frontal zone, etc.). The primary focus centers on the cases associated with tropical waves, because these waves account for the majority of all Atlantic tropical cyclones. The detection technique illustrates the ability to track these tropical disturbances from near the coast of Africa. Analysis of the pretropical cyclone (pre-TC) tracks for these cases depicts stages, related to wind speed and precipitation, in the evolution of a tropical disturbance within an easterly wave to a tropical cyclone.

1. Introduction

Although tropical cyclogenesis (TCG) is a very active area of research, it remains a highly debatable and unresolved topic. While considerable attention has been paid to tropical cyclone formation, little attention has focused on observational studies of the very early stages of TCG, otherwise referred to as the genesis stage. In the past, the early stages of TCG were unverifiable in surface observations because of the paucity of meteorological data over the tropical oceans. The advent of wide-swath scatterometers helped alleviate this issue by providing the scientific community with widespread observational surface data across the tropical basins. One such instrument is the SeaWinds scatterometer, aboard the Quick Scatterometer (QuikSCAT) satellite, which infers surface wind speed and direction. Launched in 1999, this scatterometer has encouraged various studies regarding early identification of tropical disturbances (Liu 2001; Katsaros et al. 2001; Sharp et al. 2002). These studies, although operational in intent, hypothesized the potential for SeaWinds data to be applied toward research applications (i.e., genesis stage research). The main goal of this study is to examine this potential, to develop an objective technique that will detect the early stages of TCG in the Atlantic Ocean basin using SeaWinds data.

Liu (2001), Katsaros et al. (2001), and Sharp et al. (2002) demonstrated the ability to identify tropical disturbances, that is, discrete weather systems of apparently organized convection that maintain their identity for 24 h or more but are too weak to be classified as tropical cyclones (i.e., tropical depressions, tropical storms, and/or hurricanes) by the National Hurricane Center (NHC). Each technique utilized surface wind data obtained by the SeaWinds scatterometer; however, the criteria that defined their identification method differed. Sharp et al. (2002) employed surface relative vorticity in their detection condition, whereas Liu (2001) and Katsaros et al. (2001) relied upon closed circulations apparent in the scatterometer data. Using a threshold of vorticity over a defined area, Sharp et al. (2002) identified numerous tropical disturbances and assessed whether or not they were likely to develop into tropical cyclones. Detection was based on surface structure, requiring sufficiently strong vorticity averaged over a large surface area. The early identification of surface circulations presented in these studies suggests an opportunity to detect the early stages of TCG, setting the basis for this paper.

The detection technique described herein has the potential for applications in both the scientific and operational communities. In operational applications, the detection technique can be implemented as an additional observational (not forecasting) tool. In doing so, the technique can enhance the current observing system employed to identify and monitor tropical weather systems, thereby reducing the time forecasters spend examining the Tropics for incipient systems. In research applications, identification of the early stages of TCG can enhance understanding in regions where little research has been conducted because of the lack of surface observations, prior inability to conclusively locate TC precursor disturbances, and consequently the lack of observational studies (Reasor et al. 2005) on the establishment of the initial surface vortex. This study focuses on the Atlantic basin, but the detection technique can be applied to other tropical regions, such as the Pacific basin, after adjusting the threshold values to account for regional differences in TCG mechanisms.

The ability to detect the early stages of TCG provides an opportunity to classify tropical disturbances in the Atlantic basin based on the source of initial surface cyclonic vorticity. Following categorization by Bracken and Bosart (2000), these sources include disturbances associated with 1) monsoon troughs or the intertropical convergence zone (Riehl 1954, 1979), 2) an easterly wave (Carlson 1969; Burpee 1972, 1974, 1975; Reed et al. 1977; Thorncroft and Hoskins 1994a, b), 3) a stagnant frontal zone originating in the midlatitudes (Frank 1987; Davis and Bosart 2001), 4) mesoscale convective systems (MCSs; Bosart and Sanders 1981; Ritchie and Holland 1997; Simpson et al. 1997; Bister and Emanuel 1997; Montgomery and Enagonio 1998), and 5) upper-level cutoff lows that penetrate to lower levels (Avila and Rappaport 1996). Among these, our research in the Atlantic basin affords the possibility to investigate cases associated with easterly waves. Of great interest is the prospect to examine the connection between a cold-core wave disturbance in the tropical easterlies and a warm-core tropical cyclone, which remains an unresolved issue in TCG research. Easterly waves are of great importance since approximately 63% of tropical cyclones in the Atlantic basin originate from or within African easterly waves (Avila and Pasch 1992).

Another fundamental issue with TCG is the formation of the surface vortex prior to the onset of the Wind-Induced Surface Heat Exchange intensification mechanism of Rotunno and Emanuel (1987). The generation of surface vorticity and its interaction within regions of deep cumulonimbus convection has been hypothesized in recent numerical (Hendricks et al. 2004; Montgomery et al. 2006) and observational (Reasor et al. 2005) studies to lead to the establishment of the initial surface vortex. The fine details of this process of surface vorticity organization cannot be resolved with SeaWinds data. However, SeaWinds data may resolve the general evolution of surface vorticity and provide estimates of vorticity available at the surface prior to TCG, which are needed for numerical model initialization and validation. Such models are not examined herein. We present an objective technique that can be used to identify tropical disturbances far earlier in their evolution than is typically capable with previous techniques. This technique is a critical precursor to modeling studies with realistic initial conditions.

The outline for this paper is as follows. Background information regarding the basic principles of scatterometry and the SeaWinds scatterometer is presented in section 2. In section 3, the scatterometer data and Geostationary Operational Environmental Satellite (GOES) imagery employed in this study are detailed. The methodology behind the vorticity-based detection technique, including determination of threshold values and track assessment is described in section 4. Results of the detection technique are presented in section 5, with subsections dedicated to two individual case studies. The influences of rain on QuikSCAT winds are discussed and related to the evolution of wind and rain rates in tropical disturbances in section 6. Conclusions are summarized in section 7. Overall, the detection technique proves very effective, identifying tropical disturbances associated with the early stages of TCG approximately 19–101 h before they are classified as tropical cyclones by the NHC.

2. Background

a. Principles of scatterometry

Scatterometers emit microwave pulses to the ocean surface and measure the backscattered power in the return echo. Microwaves are scattered by small-scale water waves (i.e., capillary waves), which are in near equilibrium with the wind. As a result, capillary waves react quickly to surface wind changes (i.e., wind speed and direction). Modulation of the waves alters the surface roughness of the ocean, affecting the radar cross section and, hence, the magnitude of the backscattered power received by the instrument. From the backscattered power, the normalized radar cross section or backscatter (σ°) is obtained by inverting the radar equation (Naderi et al. 1991).

Scatterometers provide multiple measurements of σ° at various azimuth angles. These measurements are organized into square bins called wind vector cells (WVCs). The method that estimates the wind velocity associated with σ° measurements is the wind retrieval process. The first step of the wind retrieval process is wind inversion. Wind inversion determines the scatterometer wind solutions that minimize an objective function, which measures the squared difference between observed and modeled backscatter (σ°). The modeled backscatter (σ°) is determined by the geophysical model function (GMF).

The GMF is an empirical nonlinear model that describes the relationship between wind velocity and σ°. This function is written as

 
formula

where u is the wind speed, χ is the relative direction (the angle between the wind direction and the look direction of the scatterometer), θ is the surface incidence angle, pol is the electromagnetic polarization, and f is the radar frequency (Naderi et al. 1991). Other factors such as temperature, ocean salinity, and foam also affect the relationship between wind and σ°, but they are assumed to be small and considered noise in the GMF (Richards 1999).

Because of near symmetry of the GMF’s dependence on the look direction, the objective function typically has several local minima associated with different wind directions. The wind solutions associated with these minima are otherwise referred to as ambiguities or aliases. These aliases are sorted according to rank, where rank 1 corresponds to the solution that has the lowest objective function value, rank 2 to the solution that has the next lowest objective function value, and so on (Naderi et al. 1991). The second step of the wind retrieval process is an ambiguity removal algorithm. Because of noise and the multiple wind solutions (i.e., ambiguities) that result from wind inversion, this step is often necessary to yield a unique wind vector solution. Correct ambiguity removal results in the selection of the ambiguity that is closest to the actual wind vector (Naderi et al. 1991). Unfortunately, ambiguity removal algorithms are prone to errors (section 2b).

b. The SeaWinds scatterometer

The QuikSCAT satellite was launched on 19 June 1999 as a “quick recovery” mission to fill the gap created by the loss of data from the National Aeronautics and Space Administration (NASA) scatterometer (NSCAT) when the Advanced Earth Observing Satellite (ADEOS) satellite lost power in June 1997. It sun-synchronously orbits Earth, with an orbital height of approximately 800 km above the equator. The orbital and recurrent periods of the QuikSCAT satellite are 101 min and 4 days, respectively. QuikSCAT swaths are 1800 km wide and cover over 90% of Earth’s surface in 24 h. Globally averaged, there are close to two samples per day, with more nearer the Poles, and less nearer the equator.

The measuring instrument aboard the QuikSCAT satellite is the SeaWinds scatterometer. It is an active microwave sensor that uses a rotating dish antenna with two conically rotating pencil beams to acquire multiple measurements of backscattered power from different viewing geometries, which are then organized into 25 km × 25 km WVCs. Herein a horizontally polarized (h pol) beam at an incidence angle of 46.25° and a vertically polarized (v pol) beam at an incidence angle of 54° trace the surface of Earth in a helical pattern, receiving backscatter measurements fore and aft the orbiting satellite (Fig. 1). The measurement geometry that results from this scanning beam configuration varies across the swath. At swath edges there is very little azimuth variation between the fore and aft beams, and measurements taken are only vertically polarized; however, in the nadir region the azimuth difference between the fore and aft beams is approximately 180° and measurements taken are both vertically and horizontally polarized. However, the measurements in the nadir region are nearly identical as a result of symmetry in the GMF. Therefore, poor viewing geometries for wind retrieval occur in the nadir region and at swath edges. The optimal region for wind retrieval is in the midswath, which is also known as the “sweet spot.” The midswath is approximately 200–700 km on either side of the satellite track.

Fig. 1.

The spacecraft and antenna geometries for SeaWinds. The outer beam is v pol and the inner beam is h pol (adapted from Weissman et al. 2003).

Fig. 1.

The spacecraft and antenna geometries for SeaWinds. The outer beam is v pol and the inner beam is h pol (adapted from Weissman et al. 2003).

The ambiguity removal algorithm used by QuikSCAT is a vector median filter (Shaffer et al. 1991). This filter is initialized with the ambiguity closest to the National Centers for Environmental Prediction (NCEP) 2.5° wind fields (Wentz and Smith 1999). Once initialized, the median filter repeatedly selects the ambiguity at each WVC that is closest to the actual wind vector until convergence is attained. The shortcoming of this approach, however, is that it is only successful if a majority of the initial ambiguities correspond to the true wind (Shaffer et al. 1991). Since the selected winds are horizontally consistent, ambiguity removal errors tend to occur in patches or lines and are often approximate direction reversals (roughly 180°).

For SeaWinds, wind retrieval is problematic for several conditions. First, the wind retrieval process is ill conditioned at the edges of swaths. At the edge, the median filter has fewer neighboring points to use. Second, wind retrieval is less accurate at low wind speeds (<4 m s−1) because of a low signal-to-noise ratio (Donelan and Pierson 1987). At these speeds, parts of the ocean surface act more like smooth reflectors than scatterers. Third, rain can greatly influence backscatter observations. In the presence of rain, the characteristics of the scatterometer signal are distorted because of backscatter off the rain, attenuation of the signal passing through the rain (Weissman et al. 2002), and modification of the surface shape by raindrop impacts (Bliven et al. 1993; Sobieski and Bliven 1995; Sobieski et al. 1999). As a result, wind directions in rain-contaminated WVCs (where the rain signal is similar or greater in magnitude to the wind signal) are often cross swath rather than the true direction.

3. Data

Ocean wind vectors are obtained from the SeaWinds scatterometer for the development region in the Atlantic basin. QuikSCAT has an average resampling time between 16 and 17 h for this area and, hence, provides infrequent temporal sampling. To provide continuity of the track and verification of tropical disturbances between the relatively sparse QuikSCAT overpasses, GOES infrared images are acquired and compiled into animations. These animations allow cloud features that are associated with the surface vorticity signatures to be tracked. Use of the conjunction of QuikSCAT and GOES is discussed in more detail in section 4b.

a. Scatterometer data

The scatterometer data that are used in this study are the “Ku2001” dataset produced by Remote Sensing Systems (RSS). The Ku2001 product is currently the most accurate GMF for most meteorological conditions (Bourassa et al. 2003). It performs far better near nadir, swath edges, and rain than either the science quality product from the Jet Propulsion Laboratory or the near real-time product from the National Oceanic and Atmospheric Administration (NOAA) National Environmental Satellite, Data, and Information Service (NESDIS). This improved vector wind retrieval algorithm provides a fully integrated stand-alone rain flag and the capability to retrieve winds up to 70 m s−1 for rain-free conditions where the high winds exist over a sufficiently large area (Wentz et al. 2001). The scatterometer winds are calibrated to equivalent neutral winds at a height of 10 m above the local mean water surface (Bourassa et al. 2003).

b. GOES imagery

GOES circles Earth in a geosynchronous orbit approximately 35 800 km above Earth’s equator. GOES-East is located at 75°W and GOES-West is located at 135°W. For our research, we utilize GOES-East because of its location over the Atlantic Ocean. The two GOES-East satellites are GOES-8 (GOES-I) and GOES-12 (GOES-M). GOES-12 is the current satellite in operation. It replaced GOES-8 on 3 April 2003. Both satellites have a three-axis, body-stabilized design, which provides significant improvements in weather imagery and atmospheric sounding information. They are equipped with a separate imager and sounder allowing simultaneous and independent imaging and sounding; however, we only focus on the imager with an infrared resolution of approximately 4 km at nadir.

GOES-8 and GOES-12 infrared images are obtained from the NOAA/NESDIS Comprehensive Large Array-Data Stewardship System (CLASS) for our 15 tropical cyclone cases during the 1999–2004 Atlantic hurricane seasons. Images are acquired approximately every 3 h and compiled into separate animations, with a backward- and forward-in-time progression. The examination of these animations was the time limiting factor for this investigation, limiting the number of cases that could be examined. Early in this process, we found that far more overpasses could be found for tropical disturbances that developed from easterly waves. We also found that such cases provided a wider range of spatial extents and averaged vorticity values.

4. Method

a. Detection technique

The vorticity-based detection technique used in this study is a variation of the method developed by Sharp et al. (2002). This technique calculates relative vorticity within the SeaWinds swaths and applies a mean vorticity threshold over a specified spatial area (Sharp et al. 2002). Different criteria are utilized than those of Sharp et al. (2002), however, permitting earlier identification of tropical disturbances.

The spatial scale for averaging vorticity within the SeaWinds swaths is a 100 km by 100 km area. Individual vorticity values are calculated from wind observations, defined by 4 (i.e., 2 × 2) adjacent scatterometer vectors, by determining the circulation around each box and then dividing by the area (Sharp et al. 2002). This method enables the vorticity to be calculated at the same spatial density as the wind observations. In each calculation a minimum of three wind vectors out of the four in a square are required (if only three wind vectors exist, the square becomes a triangle). The wind vector data we use in this approach include rain-flagged data, which are prone to ambiguity removal errors (reversal of wind direction). Incorporation of rain-flagged data can affect the vorticity calculations, resulting in noise. How this noise compares to the signal varies across the QuikSCAT vorticity-based track of small tropical disturbances and is discussed in section 6.

The detection technique consists of three components. These criteria require a spatial scale: in this case, a 100 km by 100 km area. First, the average vorticity must exceed a minimum vorticity threshold, and second the maximum rain-free wind speed must exceed a minimum wind speed threshold. The third criterion is that these conditions be met for at least 80% of the overlapping 100 by 100 km boxes centered on the vorticity points within 50 km of the vorticity points being tested. If these criteria are met, then the system under consideration is deemed a tropical disturbance, which may develop into a tropical cyclone. For this study, the 15 cases chosen were classified as tropical cyclones and, hence, are known to develop. The thresholds used here are greatly reduced from those of Sharp et al. (2002).

b. Track assessment

The combination of QuikSCAT’s 4-day repeat cycle and approximate 1.5 time daily coverage affords the Atlantic development region with infrequent temporal samplings (11–36 h), which proves problematic in regards to studies based purely on QuikSCAT. These temporal gaps may not be a problem for detection of existing tropical cyclones or tropical disturbances near classification; however, they are significant for identification of tropical disturbances associated with the early stages of TCG. Gaps between detection generate uncertainty regarding a tropical disturbance’s track and positioning in time. To provide continuity of the track, which is used to validate the vorticity signatures identified by the detection technique, the combination of GOES and QuikSCAT is necessary.

GOES infrared images provide supplementary observational guidance between the relatively sparse QuikSCAT overpasses. For the 15 cases, GOES animations are created with a backward- and forward-in-time progression. This process was highly time consuming, and was the limiting factor on the number of tracks examined. These animations allow the cloud mass associated with a tropical disturbance to be tracked, providing insight into the position and track extent of the tropical disturbance. For a large, organized cloud mass, the location of the corresponding surface vorticity signature is evident within QuikSCAT-derived relative vorticity swaths. However, through the pretropical cyclone tracks of our 15 cases, it is apparent that the cloud masses go through phases of intensification and decay (i.e., maturing and dissipating cloud masses). Categorization of these phases is dependent upon spatial size, organization, and convective features. If the cloud mass associated with a tropical disturbance is going through decay it may separate into numerous cloud clusters, making it difficult to pinpoint a specific cloud cluster and, hence, a position to use in testing the detection algorithm. The use of QuikSCAT and GOES together greatly reduces the ambiguity in determining which cloud system to track.

c. Threshold determination

The threshold values defined in our detection technique are determined using research-quality SeaWinds data for 15 tropical cyclones during the 1999–2004 Atlantic hurricane seasons. In preliminary examples we applied a speed and vorticity threshold of 4.0 m s−1 and 2.0 × 10−5 s−1, respectively. This first guess was expected to identify too many tropical disturbances. In all cases there was a surface vorticity signature with a closely collocated cloud signature at a location consistent with a believable trajectory. In the vast majority of cases (63 of 65 overpasses) there was a surface vorticity signature that was much larger than vorticity values 5°–10° away from the vorticity believed to be associated with a tropical disturbance. In two of the cases, the vorticity that might be associated with the cloud cover that later developed into a tropical cyclone was not suitably distinguishable from vorticity values 5°–10° away from that location. Therefore, it was determined (confirmed) that these threshold values were too small. Ideally, threshold values would be set using a statistic that at least combines hits (correct claims of detections) and false alarms (incorrect claims of detection). Applying such an ideal requires a reliable set of comparison data. No such dataset is available: we are working on spatial scales far too fine for reanalysis products or subjective analyses of easterly waves. In theory, larger detection thresholds will result in a reduced false-alarm rate (FAR), up to the point at which the thresholds are far too large. Our goal is to reduce the FAR (which we cannot determine) and retain a good probability of detection (POD). Based on the preliminary tests, it is clear that we can raise our thresholds and retain a good POD value. Furthermore, since the vorticity signatures believed to be associated with tropical disturbances are much larger (in value and in spatial extent) than environmental values, the FAR should be greatly reduced by increasing the threshold values.

The POD is defined as

 
formula

where H is the number of hits and M is the number of misses. Using the GOES imagery, the cloud cluster broadly associated with each of the 15 classified tropical cyclones is traced back in time. In QuikSCAT overpasses, a “hit” is a vorticity signature that fulfills the detection technique’s criteria within close proximity (175 km) to the identified relevant cloud cluster center, and a “miss” is a vorticity signal that does not meet the detection technique’s criteria within this close proximity. The POD score measures the ability of our technique to accurately identify tropical disturbances in the correct locations. A score of 1 indicates perfect detection (all vorticity signatures in the QuikSCAT overpasses are hits), whereas a score of 0 represents negligible detection (all vorticity signatures in the QuikSCAT overpasses are misses).

To examine this method, a POD plot is produced that assesses the contributions from both wind speed and vorticity thresholds in regards to our preliminary example (Fig. 2). This plot illustrates that low threshold values result in a higher probability of detection (albeit more false alarms), whereas, high thresholds result in a lower probability of detection. A POD score of 1 is most desirable since it represents perfect detection; however, the test cases include two examples that are indistinguishable from noise. Through analysis, the 96% POD contour is chosen based on its large gradient (high sensitivity area) in Fig. 2, as well as its successful nondetection of the two cases that we categorize as false alarms. Threshold values associated with this contour include a vorticity and wind speed threshold of 5.0 × 10−5 s−1 and 6.3 m s−1, respectively. Utilization of these values within our detection technique shows that 62 wind fields sampled by the overpasses meet our criteria.

Fig. 2.

The POD graph of TC precursors. Lower threshold values result in a higher probability of detection and increased false alarms. Higher threshold values result in detection of stronger systems (more misses) and, hence, fewer false alarms.

Fig. 2.

The POD graph of TC precursors. Lower threshold values result in a higher probability of detection and increased false alarms. Higher threshold values result in detection of stronger systems (more misses) and, hence, fewer false alarms.

5. Results

Results for the 15 tropical cyclones during the 1999–2004 Atlantic hurricane seasons are listed in Table 1. These systems are chosen because their coverage by QuikSCAT is adequate for reasonable study in the Atlantic basin. Ten of the 15 tropical cyclones originate as tropical waves off the coast of Africa (i.e., African easterly waves). These include Floyd (1999), Debby (2000), Nadine (2000), Jerry (2001), Dolly (2002), Danny (2003), Isabel (2003), Juan (2003), Nicholas (2003), and Alex (2004). The other five cases originate from sources other than tropical waves, such as upper-level cutoff lows and stagnant frontal zones. These include Florence (2000), Michael (2000), Karen (2001), Noel (2001), and Gustav (2002).

Table 1.

Results for 15 tropical cyclones during the 1999–2004 Atlantic hurricane seasons. The last column signifies the hours elapsed between the NHC initial classification and our earliest tropical disturbance identification (i.e., the tracking time).

Results for 15 tropical cyclones during the 1999–2004 Atlantic hurricane seasons. The last column signifies the hours elapsed between the NHC initial classification and our earliest tropical disturbance identification (i.e., the tracking time).
Results for 15 tropical cyclones during the 1999–2004 Atlantic hurricane seasons. The last column signifies the hours elapsed between the NHC initial classification and our earliest tropical disturbance identification (i.e., the tracking time).

Tropical disturbances associated with the early stages of TCG are found for these cases within a range of 19–101 h before classification as tropical cyclones by the NHC (Table 1). The average tracking time for these systems is approximately 58 h, where tracking time is defined as the time elapsed between the NHC initial classification and our earliest tropical disturbance identification. Some examples of the technique in identifying the early stages of TCG are illustrated in Figs. 3a–d. These figures show QuikSCAT-derived relative vorticity overlaid with solid black contours, signifying the locations where the detection technique’s criteria are met within 175 km from the cloud cluster center in the associated GOES infrared image.

Fig. 3.

Examples of the early stages of TCG that are identified by the vorticity-based detection technique. The background color represents QuikSCAT-derived relative vorticity, with dark green representing all vorticity values greater than 5.0 × 10−5 s−1. The black solid lines signify the locations where the detection technique’s criteria are met within 75 km from the cloud cluster center in the associated GOES infrared image. (a) Floyd: 46 h before classification as a tropical depression. The vorticity signature shown is associated with the easterly wave that spawns Floyd. (b) Nicholas: 64 h before classification as a tropical depression. The vorticity signature shown is associated with the easterly wave that produces Nicholas. (c) Michael: 38 h before classification as a subtropical depression. The vorticity signature shown is associated with the upper-level cold low that interacted with a stationary front to create Michael. (d) Noel: 26 h before classification as a subtropical storm. The vorticity signature shown is associated with the nontropical occluded low that spawns Noel. Each example illustrates an apparent surface circulation.

Fig. 3.

Examples of the early stages of TCG that are identified by the vorticity-based detection technique. The background color represents QuikSCAT-derived relative vorticity, with dark green representing all vorticity values greater than 5.0 × 10−5 s−1. The black solid lines signify the locations where the detection technique’s criteria are met within 75 km from the cloud cluster center in the associated GOES infrared image. (a) Floyd: 46 h before classification as a tropical depression. The vorticity signature shown is associated with the easterly wave that spawns Floyd. (b) Nicholas: 64 h before classification as a tropical depression. The vorticity signature shown is associated with the easterly wave that produces Nicholas. (c) Michael: 38 h before classification as a subtropical depression. The vorticity signature shown is associated with the upper-level cold low that interacted with a stationary front to create Michael. (d) Noel: 26 h before classification as a subtropical storm. The vorticity signature shown is associated with the nontropical occluded low that spawns Noel. Each example illustrates an apparent surface circulation.

Although we have cases for a variety of sources, our primary focus centers on those cases associated with easterly waves. Therefore, individual case studies are presented in subsequent subsections for Isabel (2003) and Debby (2000). These cases are chosen because of the ability of our detection technique to track the tropical disturbances associated with these tropical cyclones back to the coast of Africa where they originate as easterly waves.

The development of a tropical cyclone from an easterly wave is clearly evident in the GOES images associated with pretropical cyclone (pre TC) tracks of the easterly wave cases. This evolution is depicted through comparison of the wind and rain signatures in the areas where the detection technique’s criteria are fulfilled within 175 km from the cloud cluster center. Based upon the relationship between the wind and rain signatures, the evolution is subdivided into stages (section 6).

a. Isabel (2003)

At 1856 UTC 1 September 2003, an initial vorticity signature associated with the easterly wave that spawns Tropical Cyclone Isabel is identified off the coast of Africa (Figs. 4a and 5a). The wave continues to progress westward over the next several days and gradually becomes more organized (Figs. 4b–d). Vorticity signatures associated with the wave become more consolidated and the surrounding wind pattern begins a counterclockwise rotation (Figs. 5b–d). At 1918 UTC 4 September 2003 (Fig. 5e), an apparent weak surface circulation is exhibited in the vorticity signature, with several wind observations of 10–15 m s−1 to the south of the signature. Such a circulation is not clearly evident in the associated GOES infrared image, which is still broad and disorganized (Fig. 4e). On 5 September the surface circulation is stronger and more pronounced, with numerous wind observations of 10–15 m s−1 to the north of the signature (Figs. 5f,g). GOES infrared images at these same times also exhibit cyclonic circulation, as well as organized convection (Figs. 4f,g). As a result of this adequate organized convection, satellite-based Dvorak estimates begin at 0000 UTC 5 September 2003 (Beven and Cobb 2003). At 0000 UTC 6 September 2003, the NHC classifies the tropical disturbance as a tropical depression (Beven and Cobb 2003). This track is illustrated in Fig. 6, where the triangles represent the associated QuikSCAT vorticity images, and the number above or below each triangle signifies the tracking time for that image.

Fig. 4.

GOES infrared images associated with pre-TC Isabel (2003) at (a) 1745 UTC 1 Sep 2003, (b) 0845 UTC 2 Sep 2003, (c) 0845 UTC 3 Sep 2003, (d) 2045 UTC 3 Sep 2003, (e) 2045 UTC 4 Sep 2003, (f) 0845 UTC 5 Sep 2003, and (g) 2045 UTC 5 Sep 2003 (courtesy of NOAA/NESDIS/CLASS; see online at www.class.noaa.gov).

Fig. 4.

GOES infrared images associated with pre-TC Isabel (2003) at (a) 1745 UTC 1 Sep 2003, (b) 0845 UTC 2 Sep 2003, (c) 0845 UTC 3 Sep 2003, (d) 2045 UTC 3 Sep 2003, (e) 2045 UTC 4 Sep 2003, (f) 0845 UTC 5 Sep 2003, and (g) 2045 UTC 5 Sep 2003 (courtesy of NOAA/NESDIS/CLASS; see online at www.class.noaa.gov).

Fig. 5.

QuikSCAT vorticity images associated with pre-TC Isabel (2003) and the corresponding tracking times: (a) 1856 UTC 1 Sep 2003, 101 h; (b) 0739 UTC 2 Sep 2003, 88 h; (c) 0714 UTC 3 Sep 2003, 65 h; (d) 1944 UTC 3 Sep 2003, 52 h; (e) 1918 UTC 4 Sep 2003, 29 h; (f) 0803 UTC 5 Sep 2003, 16 h; and (g) 2032 UTC 5 Sep 2003, 3 h.

Fig. 5.

QuikSCAT vorticity images associated with pre-TC Isabel (2003) and the corresponding tracking times: (a) 1856 UTC 1 Sep 2003, 101 h; (b) 0739 UTC 2 Sep 2003, 88 h; (c) 0714 UTC 3 Sep 2003, 65 h; (d) 1944 UTC 3 Sep 2003, 52 h; (e) 1918 UTC 4 Sep 2003, 29 h; (f) 0803 UTC 5 Sep 2003, 16 h; and (g) 2032 UTC 5 Sep 2003, 3 h.

Fig. 6.

Track of the tropical disturbance associated with pre-TC Isabel (2003). Triangles represent the associated QuikSCAT images, and the number above or below each triangle signifies the tracking time for that image. The number “0” represents when the NHC classified Isabel as a tropical depression.

Fig. 6.

Track of the tropical disturbance associated with pre-TC Isabel (2003). Triangles represent the associated QuikSCAT images, and the number above or below each triangle signifies the tracking time for that image. The number “0” represents when the NHC classified Isabel as a tropical depression.

b. Debby (2000)

Debby develops from a strong African easterly wave (Pasch 2000). An initial vorticity signature associated with this wave is observed at 1859 UTC 15 August 2000 (Figs. 7a and 8a). As the wave propagates westward, the winds gradually develop a cyclonic curvature and a weak surface cyclonic circulation becomes evident on 17 August (Figs. 7b–d and 8b–d). The NOAA/NCEP Tropical Prediction Center (TPC) Tropical Analysis and Forecast Branch (TAFB) observe this broad area of low pressure and are unable to give the disturbance Dvorak classification since there is insufficient curvature in the associated bands of deep convection (Pasch 2000). Therefore, the system is characterized as “too weak to classify” and is monitored (Pasch 2000). Through 18 August the surface circulation continues to become more organized, illustrating a tighter cyclonic rotation and higher wind speeds of 5–15 m s−1 (Figs. 8e,f). Such characteristics are not as evident in the associated GOES infrared images, which still show the cloud pattern of the disturbance to be broad and poorly organized (Figs. 7e,f). An increase in curvature of the convective bands prompted Dvorak classification of the disturbance by the TAFB at 1145 UTC 18 August 2000 (Pasch 2000). The disturbance continues to track westward at approximately 10 m s−1, becoming more organized (Figs. 7g and 8g). Approximately 900 nautical miles east of the Windward Islands the cloud pattern associated with the disturbance becomes consolidated around a well-defined center and is categorized by the NHC as a tropical depression at 1800 UTC 19 August 2000 (Pasch 2000). This track is illustrated in Fig. 9.

Fig. 7.

GOES infrared images associated with pre-TC Debby (2000): (a) 2045 UTC 15 Aug 2000, (b) 0845 UTC 16 Aug 2000, (c) 2045 UTC 16 Aug 2000, (d) 2045 UTC 17 Aug 2000, (e) 0845 UTC 18 Aug 2000, (f) 2045 UTC 18 Aug 2000, and (g) 0845 UTC 19 Aug 2000 (provided by NOAA/NESDIS/CLASS).

Fig. 7.

GOES infrared images associated with pre-TC Debby (2000): (a) 2045 UTC 15 Aug 2000, (b) 0845 UTC 16 Aug 2000, (c) 2045 UTC 16 Aug 2000, (d) 2045 UTC 17 Aug 2000, (e) 0845 UTC 18 Aug 2000, (f) 2045 UTC 18 Aug 2000, and (g) 0845 UTC 19 Aug 2000 (provided by NOAA/NESDIS/CLASS).

Fig. 8.

QuikSCAT vorticity images associated with pre-TC Debby (2000) and the corresponding tracking times: (a) 1859 UTC 15 Aug 2000, 95 h; (b) 0742 UTC 16 Aug 2000, 82 h; (c) 2014 UTC 16 Aug 2000, 70 h; (d) 1950 UTC 17 Aug 2000, 46 h; (e) 0833 UTC 18 Aug 2000, 33 h; (f) 2104 UTC 18 Aug 2000, 21 h; and (g) 0808 UTC 19 Aug 2000, 10 h.

Fig. 8.

QuikSCAT vorticity images associated with pre-TC Debby (2000) and the corresponding tracking times: (a) 1859 UTC 15 Aug 2000, 95 h; (b) 0742 UTC 16 Aug 2000, 82 h; (c) 2014 UTC 16 Aug 2000, 70 h; (d) 1950 UTC 17 Aug 2000, 46 h; (e) 0833 UTC 18 Aug 2000, 33 h; (f) 2104 UTC 18 Aug 2000, 21 h; and (g) 0808 UTC 19 Aug 2000, 10 h.

Fig. 9.

Same as in Fig. 6, but for pre-TC Debby (2000).

Fig. 9.

Same as in Fig. 6, but for pre-TC Debby (2000).

6. Discussion

Results from the vorticity-based detection technique prove very effective in identification of tropical disturbances; however, the principal concern with this technique regards the ambiguity removal errors associated with rain-flagged data. These selection errors are evident for the majority of the 15 cases, but the cases more seriously affected are those originating in the Tropics (i.e., easterly waves). Disturbances connected with this origin exhibit lower wind speeds and high moisture content, contributing to the significant number of rain flags and, hence, ambiguity removal errors. In contrast, cases that develop from sources other than tropical waves, such as upper-level cutoff lows and stagnant frontal zones, are generally unaffected by the negative influences of rain (i.e., selection errors). Systems associated with such sources have subtropical origins and therefore exhibit higher wind speeds and reduced moisture content. A key example of a subtropical system that illustrates such features is the disturbance associated with pre-TC Noel (2001) on 2 November (Fig. 3d).

Incorporation of these selection errors within the technique significantly impacts the vorticity calculation. As previously mentioned, average vorticity is calculated within a 100 km by 100 km area using individual vorticity values, which are determined from wind observations via the circulation theorem (section 4a). Therefore, if the wind observations used to calculate vorticity on the outer bounds of the spatial domain include ambiguity removal errors, the average vorticity is affected by these errors; ambiguity selection errors on the interior of the domain have no influence on the vorticity calculation. Because of the size and shape of our spatial domain for vorticity calculations, the locations where these selection errors (i.e., incorrectly selected ambiguities) influence the averaged vorticity are approximately 100–140 km from the selection errors.

Cases associated with easterly waves can be considerably affected by the ambiguity selection errors associated with rain-flagged data. Results from the case studies in section 5 demonstrate this problem. In these two case studies, the importance of rain-related issues on wind speeds and direction varies throughout the swaths. The relationship between these signals (i.e., wind speed and rain) has been previously characterized (Weissman et al. 2002; Draper and Long 2004). In these cases suspect vorticity values are found at the expected distance from suspect wind vectors (vectors that are rain flagged and greatly different from surrounding vectors).

The amount of backscattered power that is received by the scatterometer is a function of wind speed and rain characteristics (Weissman et al. 2002). If the wind signal is larger than that of the rain signal, then rain insignificantly affects the backscatter signal. However, if the rain signal is larger than that of the wind signal, then rain influences the backscattered power. The influence of rain distorts the scatterometer signal, potentially creating wind direction reversals in the data that affect the average vorticity calculation. In areas strongly dominated by rain, the scatterometer directions are perpendicular to the nadir track, regardless of the actual wind direction.

Based upon the relationship between wind and rain signatures in the areas where the detection technique’s criteria are met (within 175 km from the cloud cluster center), the pre-TC tracks of each easterly wave case are subdivided into stages. These stages consist of an initial, intermediate, and near-TC phase. The initial stage corresponds to the earliest vorticity signatures identified by the detection technique, the near-TC stage is associated with the vorticity signatures detected directly prior to NHC classification, and the intermediate stage constitutes the vorticity signatures between the initial and near-TC stages. Each of these phases is discussed in detail in subsequent sections (sections 6ac). These stages depict the transition of a tropical disturbance within an easterly wave to a tropical cyclone, and are further discussed in section 6d.

a. The initial stage

The vorticity signatures that are identified by the detection technique are small in spatial extent and very weak. Examples of these diminutive, weak signatures are illustrated in Figs. 5a and 8a. As seen in these images, the initial vorticity signatures correspond with rain-flagged wind vectors and few large ambiguity removal errors (i.e., wind direction reversals). This situation implies that the wind signal is larger than the rain signal (Weissman et al. 2002). Therefore, any vorticity modification that occurs as a result of selection errors is small when compared with the signal.

b. The intermediate stage

Relative to the initial stage, the vorticity signatures identified in the intermediate stage are stronger and have a larger spatial size (Figs. 5b–d and 8b,c). The areas where the detection technique’s criteria are fulfilled (black solid contours) are inundated with numerous rain-flagged wind vectors and ambiguity removal errors, implying that the vertically integrated rain rates have increased relative to the initial stage. The rain-flagged wind vectors are comparatively large in magnitude and exhibit an across swath direction, signifying rain contamination. This situation implies that the rain signal has grown to dominate the wind signal. Medium to strong tropical storms also have this characteristic. Thus, rain-related vorticity modification is equivalent or large when compared with the signal. In these instances, several of the areas meeting the detection criteria appear to be a result of ambiguity selection errors (Figs. 5b–d and 8b,c). For example, the QuikSCAT vorticity image associated with pre-TC Isabel at 1944 UTC 3 September 2003 depicts large and small areas that meet the criteria (Fig. 5d). The larger area contains what appears to be the correct location, depicting a developing cyclonic wind curvature pattern; whereas, the smaller location area is a by-product of selection errors. In regards to this stage, the detection technique is pinning down the general area of the tropical disturbance and, although some identified signatures are by-products of selection errors, the detection technique is effective. In general, the intermediate stage is the development phase of the pre-TC tracks. Reasoning behind this stems from the beginning of a cyclonic circulation in the surrounding wind pattern and the increase of vertically integrated rain rates.

c. The near-TC stage

Vorticity signatures associated with the near-TC stage occur approximately 1–2 days prior to NHC classification. These signatures are very organized and have a large spatial size (Figs. 5e–g and 8d–g). A large-scale cyclonic circulation is present for each signature with strong wind speeds (strong in relation to the wind speeds in the initial and intermediate stages). The near-TC vorticity signatures correspond to rain-flagged wind vectors and few ambiguity removal errors. This situation implies that the wind signal dominates the rain signal. Therefore, any vorticity modification that occurs as a result of selection errors is small when compared with the signal. Relative to the intermediate stage, the near-TC stage, with less rain-related problems, indicates that either the rain rate has been reduced (which we have not examined) or that the wind speeds have increased, the latter of which appears to be true.

d. Pre-TC track transition

The vorticity signal-to-noise pattern associated with the QuikSCAT vorticity-based tracks of the two individual case studies clearly depict the transition of a tropical disturbance within an easterly wave to a tropical cyclone and the phases associated with this transition (initial, intermediate, and near-TC stage). The initial stage is associated with very weak vorticity signatures that have small spatial scales. These initial signals are often related to horizontal shear embedded in eastward winds and are insignificantly affected by rain. The intermediate stage constitutes vorticity signatures that are greatly influenced by rain and rain-related problems, as well as the beginning of a large-scale cyclonic circulation. The intermediate stage is of significant importance to the evolution of a tropical disturbance within an easterly wave to a tropical cyclone since it represents a development stage. As a result of this growth, the vorticity signatures corresponding to the near-TC stage are highly organized, with large spatial sizes, increased wind speeds, and large-scale cyclonic circulations. The surface signatures of these tropical disturbances, as well as the cloud pattern, continue to strengthen and organize.

e. Future improvements

Scatterometer design, as well as wind retrieval algorithms, could be improved upon to better account for rain. The SeaWinds scatterometer uses the Ku-band and large incidence angles, both of which result in more sensitivity to rain. The pointwise wind retrieval method is also substantially affected by rain. However, if the incidence angles are reduced and a different frequency (e.g., C band) or combination of frequencies (e.g., Ku and C band) is used, then rain would be less of an issue. The use of a fieldwise wind retrieval method would also prove beneficial, reducing rain-related problems (i.e., ambiguity removal errors). Unlike pointwise wind retrieval, which calculates the wind at each WVC individually, fieldwise wind retrieval determines the wind for an entire region of WVCs, acknowledging that a correlation exists between each wind vector in a field of cells (Richards 1999). Overall, the addition of a radiometer [e.g., the Advanced Microwave Scanning Radiometer (AMSR)] on the same platform as the SeaWinds scatterometer would greatly improve rain-impact flags and rain corrections.

7. Conclusions

A vorticity-based detection technique is developed to identify and monitor tropical disturbances associated with the early stages of TCG in the Atlantic basin. The identification of tropical disturbances for tuning of the detection technique is based on visual inspection of GOES infrared images, as well as surface structure. Tropical systems that developed into tropical cyclones were tracked backward in time using these two types of observations. Consequently, only disturbances that grew into tropical cyclones were examined. From a sampling of 15 tropical cyclones during the 1999–2004 Atlantic hurricane seasons, the technique identifies tropical disturbances approximately 19–101 h before classification by the NHC. Herein the minimum tracking time is associated with pre-TC Karen (2001); whereas, the maximum tracking time is associated with pre-TC Jerry (2001) and pre-TC Isabel (2003). Smaller tracking times are associated with systems that develop from sources other than easterly waves, such as upper-level cutoff lows and stagnant frontal zones. The shorter tracking times stem from land interference, which QuikSCAT lacks the ability to resolve since it only provides observations over water. Larger tracking times correspond to cases that originate from African easterly waves. Focus is primarily concentrated on such cases, because approximately 63% of all Atlantic tropical cyclones are associated with tropical waves. With a means of demonstrating a low false-alarm rate, and with automation, this technique would be useful to operational forecasters.

Overall results for the 15 cases prove very successful. Therefore, the vorticity-based detection technique described herein is an effective tool in identifying and monitoring tropical disturbances in the genesis stage. This technique has not been developed for forecast applications because the lack of high-quality comparison data prevented the determination of a false-alarm rate. For easterly wave cases, the detection technique has the ability to track tropical disturbances from near the coast of Africa. The pre-TC tracks of these cases depict the evolution of a tropical disturbance within an easterly wave to a tropical cyclone and stages associated with this transition. These stages consist of an initial, intermediate, and near-TC phase, which are related to spatial extent, wind speed, and precipitation characteristics. Although there is room for improvements regarding scatterometer design and wind retrieval algorithms to better account for the influences of rain, it is reasonable to assume that this technique could be useful to scientific and operational communities.

Acknowledgments

Support for the scatterometer research came from the NASA OSU SeaWinds project, the NASA OVWST, and the NASA TCSP mission. QuikSCAT data are produced by Remote Sensing Systems and sponsored by the NASA OVWST. (Data were available online at www.remss.com.) GOES images are provided by NOAA/NESDIS/CLASS (and were available online at www.class.noaa.gov). COAPS receives its base funding from NOAA/CDEP.

REFERENCES

REFERENCES
Avila
,
L. A.
, and
R.
Pasch
,
1992
:
Atlantic tropical systems of 1991.
Mon. Wea. Rev.
,
120
,
2688
2696
.
Avila
,
L. A.
, and
E. N.
Rappaport
,
1996
:
Atlantic hurricane season of 1994.
Mon. Wea. Rev.
,
124
,
1558
1578
.
Beven
,
J.
, and
H.
Cobb
,
2003
:
Tropical cyclone report hurricane Isabel 6-19 September 2003. NCEP Rep., 24 pp. [Available online at www.nhc.noaa.gov/2003isabel.shtml?.]
.
Bister
,
M.
, and
K. A.
Emanuel
,
1997
:
The genesis of Hurricane Guillermo: TEXMEX analyses and a modeling study.
Mon. Wea. Rev.
,
125
,
2662
2682
.
Bliven
,
L. F.
,
H.
Branger
,
P. W.
Sobieski
, and
J. P.
Giovanageli
,
1993
:
An analysis of scatterometer returns from a water agitated by artificial rain.
Int. J. Remote Sens.
,
14
,
2315
2329
.
Bosart
,
L. F.
, and
F.
Sanders
,
1981
:
The Johnstown flood of July 1977: A long-lived convective system.
J. Atmos. Sci.
,
38
,
1616
1642
.
Bourassa
,
M. A.
,
D. M.
Legler
, and
J. J.
O’Brien
,
2003
:
Scatterometry data sets: High quality winds over water. Advances in the Applications of Marine Climatology—The Dynamic Part of the WMO Guide to the Applications of Marine Climatology, JCOMM Tech. Rep. 13, WMO/TD-1081, 159–174
.
Bracken
,
W. E.
, and
L. F.
Bosart
,
2000
:
The role of synoptic-scale flow during tropical cyclogenesis over the North Atlantic Ocean.
Mon. Wea. Rev.
,
128
,
353
376
.
Burpee
,
R. W.
,
1972
:
The origin and structure of easterly waves in the lower troposphere in North Africa.
J. Atmos. Sci.
,
29
,
77
90
.
Burpee
,
R. W.
,
1974
:
Characteristics of North African easterly waves during the summers of 1968 and 1969.
J. Atmos. Sci.
,
31
,
1556
1570
.
Burpee
,
R. W.
,
1975
:
Some features of synoptic-scale waves based on a compositing analysis of GATE data.
Mon. Wea. Rev.
,
103
,
921
925
.
Carlson
,
T. N.
,
1969
:
Synoptic histories of three African disturbances that developed into Atlantic hurricanes.
Mon. Wea. Rev.
,
97
,
256
276
.
Davis
,
C. A.
, and
L. F.
Bosart
,
2001
:
Numerical simulations of the genesis of Hurricane Diana (1984). Part I: Control simulation.
Mon. Wea. Rev.
,
129
,
1859
1881
.
Donelan
,
M. A.
, and
W. J.
Pierson
,
1987
:
Radar scattering and equilibrium ranges in wind generated waves with application to scatterometry.
J. Geophys. Res.
,
92
,
4971
5030
.
Draper
,
D. W.
, and
D. G.
Long
,
2004
:
Simultaneous wind and rain retrieval using SeaWinds data.
IEEE Trans. Geosci. Remote Sens.
,
42
,
1411
1423
.
Frank
,
W. M.
,
1987
:
Tropical cyclone formation.
A Global View of Tropical Cyclones, R. L. Elsberry, Ed., Office of Naval Research, 53–90
.
Hendricks
,
E. A.
,
M. T.
Montgomery
, and
C. A.
Davis
,
2004
:
The role of “vortical” hot towers in the formation of Tropical Cyclone Diana (1984).
J. Atmos. Sci.
,
61
,
1209
1232
.
Katsaros
,
K. B.
,
E. B.
Forde
,
P.
Chang
, and
W. T.
Liu
,
2001
:
QuikSCAT’s SeaWinds facilitates early identification of tropical depressions in 1999 hurricane season.
Geophys. Res. Lett.
,
28
,
1043
1046
.
Liu
,
W. T.
,
2001
:
Wind over troubled water.
Backscatter
,
12
,
10
14
.
Montgomery
,
M. T.
, and
J.
Enagonio
,
1998
:
Tropical cyclogenesis via convectively forced vortex Rossby waves in a three-dimensional quasigeostrophic model.
J. Atmos. Sci.
,
55
,
3176
3207
.
Montgomery
,
M. T.
,
M. E.
Nicholls
,
T. A.
Cram
, and
A. B.
Saunders
,
2006
:
A vortical hot tower route to tropical cyclogenesis.
J. Atmos. Sci.
,
63
,
355
386
.
Naderi
,
F. M.
,
M. H.
Freilich
, and
D. G.
Long
,
1991
:
Spaceborne radar measurements of wind velocity over the ocean—An overview of the NSCAT scatterometer system.
Proc. IEEE
,
79
,
850
866
.
Pasch
,
R. J.
,
2000
:
Tropical cyclone report hurricane Debby 19–24 August 2000. NCEP Rep., 10 pp. [Available online at www.nhc.noaa.gov/2000debby.html.]
.
Reasor
,
P. D.
,
M. T.
Montgomery
, and
L.
Bosart
,
2005
:
Mesoscale observations of the genesis of Hurricane Dolly (1996).
J. Atmos. Sci.
,
62
,
3151
3171
.
Reed
,
R. J.
,
D. C.
Norquist
, and
E. E.
Recker
,
1977
:
The structure and properties of African wave disturbances as observed during Phase III of GATE.
Mon. Wea. Rev.
,
105
,
317
333
.
Richards
,
S. L.
,
1999
:
A field-wise retrieval algorithm for SeaWinds. M.S. thesis, Dept. of Electrical and Computer Engineering, Brigham Young University, 148 pp
.
Riehl
,
H.
,
1954
:
Tropical Meteorology.
McGraw-Hill, 392 pp
.
Riehl
,
H.
,
1979
:
Climate and Weather in the Tropics.
Academic Press, 611 pp
.
Ritchie
,
E. A.
, and
G. H.
Holland
,
1997
:
Scale interactions during the formation of Typhoon Irving.
Mon. Wea. Rev.
,
125
,
1377
1396
.
Rotunno
,
R.
, and
K. A.
Emanuel
,
1987
:
An air–sea interaction theory for tropical cyclones. Part II: Evolutionary study using a nonhydrostatic axisymmetric numerical model.
J. Atmos. Sci.
,
44
,
542
561
.
Shaffer
,
S. J.
,
R. S.
Dunbar
,
S. V.
Hsiao
, and
D. G.
Long
,
1991
:
A median-filter-based ambiguity removal algorithm for NSCAT.
IEEE Trans. Geosci. Remote Sens.
,
29
,
167
174
.
Sharp
,
R. J.
,
M. A.
Bourassa
, and
J. J.
O’Brien
,
2002
:
Early detection of tropical cyclones using SeaWinds-derived vorticity.
Bull. Amer. Meteor. Soc.
,
83
,
879
889
.
Simpson
,
J.
,
E.
Ritchie
,
G. J.
Holland
,
J.
Halverson
, and
S.
Stewart
,
1997
:
Mesoscale interactions in tropical cyclone genesis.
Mon. Wea. Rev.
,
125
,
2643
2661
.
Sobieski
,
P.
, and
L. F.
Bliven
,
1995
:
Analysis of high speed images of raindrop splash products and Ku-band scatterometer returns.
Int. J. Remote Sens.
,
16
,
2721
2726
.
Sobieski
,
P.
,
C.
Craeye
, and
L. F.
Bliven
,
1999
:
Scatterometric signatures of multivariate drop impacts on fresh and salt water surfaces.
Int. J. Remote Sens.
,
20
,
2149
2166
.
Thorncroft
,
C. D.
, and
B. J.
Hoskins
,
1994a
:
An idealized study of African easterly waves. Part I: A linear view.
Quart. J. Roy. Meteor. Soc.
,
120
,
953
982
.
Thorncroft
,
C. D.
, and
B. J.
Hoskins
,
1994b
:
An idealized study of African easterly waves. Part II: A nonlinear view.
Quart. J. Roy. Meteor. Soc.
,
120
,
983
1015
.
Weissman
,
D. E.
,
M. A.
Bourassa
, and
J.
Tongue
,
2002
:
Effects of rain rate and wind magnitude on SeaWinds scatterometer wind speed errors.
J. Atmos. Oceanic Technol.
,
19
,
738
746
.
Weissman
,
D. E.
,
M. A.
Bourassa
,
J.
Tongue
, and
J. J.
O’Brien
,
2003
:
Calibrating the QuikSCAT/SeaWinds radar for measuring rain over water.
IEEE Trans. Geosci. Remote Sens.
,
41
,
2814
2820
.
Wentz
,
F. J.
, and
D. K.
Smith
,
1999
:
A model function for the ocean-normalized radar cross section at 14 GHz derived from NSCAT observations.
J. Geophys. Res.
,
104
,
11499
11514
.
Wentz
,
F. J.
,
D. K.
Smith
,
C. A.
Mears
, and
C. L.
Gentemann
,
2001
:
Advanced algorithms for QuikSCAT and SeaWinds/AMSR. Proc. Geosci. Remote Sens. Symp. (IGARSS’01), Sydney, Australia, IEEE, 1079–1081
.

Footnotes

Corresponding author address: Mark A. Bourassa, Center for Ocean–Atmospheric Prediction Studies (COAPS), The Florida State University, Tallahassee, FL 32306-2840. Email: bourassa@coaps.fsu.edu