The landfall of Hurricane Katrina (2005) near New Orleans, Louisiana, on 29 August 2005 will be remembered as one of the worst natural disasters in the history of the United States. By comparison, the extratropical transition (ET) of the system as it accelerates poleward over the following days is innocuous and the system weakens until its eventual demise off the coast of Greenland. The extent of Katrina’s perturbation of the midlatitude flow would appear to be limited given the lack of reintensification or downstream development during ET. However, the slow progression of a strong upper-tropospheric warm pool across the North Atlantic Ocean in the week following Katrina’s landfall prompts the question of whether even a nonreintensifying ET event can lead to significant modification of the midlatitude flow. Analysis of Hurricane Katrina’s outflow layer after landfall suggests that it does not itself make up the long-lived midlatitude warm pool. However, the interaction between Katrina’s anticyclonic outflow and an approaching baroclinic trough is shown to establish an anomalous southwesterly conduit or “freeway” that injects a preexisting tropospheric warm pool over the southwestern United States into the midlatitudes. This warm pool reduces predictability in medium-range forecasts over the North Atlantic and Europe while simultaneously aiding in the development of Hurricanes Maria and Nate. The origin of the warm pool is shown to be the combination of anticyclonic upper-level features generated by eastern Pacific Hurricane Hilary and the south Asian anticyclone (SAA). The hemispheric nature of the connections involved with the development of the warm pool and its injection into the extratropics has an impact on forecasting, since the predictability issue associated with ET in this case involves far more than the potential reintensification of the transitioning system itself.
Hurricane Katrina (2005) has left a devastating legacy in the United States as one of the worst natural disasters in the country’s history and the most costly hurricane ever [see McTaggart-Cowan et al. (2007), hereafter referred to as Part I, for an analysis of Hurricane Katrina’s life cycle]. Following an initial landfall on the southern Florida peninsula as a category-1 hurricane [wind speeds of 34 m s−1; Simpson (1974)], Katrina intensifies rapidly [winds increasing from 36 to 77 m s−1 (category 5) in 48 h] as it churns slowly across the Gulf of Mexico on an anticyclonically curved track that guides it toward its extremely damaging second landfall as a category-3 storm (esimated winds of 57 m s−1) near Buras, Louisiana (Fig. 1). Shortly thereafter, Katrina makes its third and final landfall on the Louisiana–Mississippi border and the left eyewall of the hurricane strikes the city of New Orleans. High winds and storm surge breach critical levies, resulting in extensive flooding throughout the city. The death toll currently stands at over 1000, and insured damages are estimated between $38 and $46 billion (U.S. dollars; Churney 2006).
After its final landfall, Katrina begins to weaken rapidly, but produces hurricane-force winds as far inland as Laurel, Mississippi. As the storm undergoes a rapid extratropical transition [ET; Fig. 2; see Jones et al. (2003) for a review of ET research], it accelerates northeastward ahead of a deepening upper-level trough; however, the wind speeds around the remnant system decrease monotonically with time and no significant reintensification of the storm is recorded. The postlandfall evolution of Hurricane Katrina’s upper-level outflow anticyclone is markedly different from that of the lower-level vortex, and forms the basis of this study. The proximity of the outflow layer to the upstream trough creates an anomalous conduit/freeway that injects a preexisting upper-level warm pool (anticyclone) into the extratropics. A number of recent ET studies have demonstrated the influence of reintensifying storms on planetary-scale circulations (e.g., Anwender et al. 2006; Harr et al. 2006; Hart 2006); however, this research addresses the following question: “Can an ET event that does not involve reintensification exert a significant influence on the midlatitude flow?”
The effects of ET on predictability have been documented by Hoskins and Berrisford (1988), Morgan (2004), Harr (2004), among others. Reduced forecast confidence is usually associated with the redevelopment of the tropical vortex as an extratropical storm, and is maximized on a 2–6-day time scale following the onset of ET (Morgan 2004). Forecast uncertainty arising from downstream development triggered by the reintensifying system has also been identified as an important result of ETs that enter a deepening phase following transition (Hoskins and Berrisford 1988). The potential for nonreintensifying systems—such as Hurricane Katrina—to cause significant predictability reductions lies in the interaction of the remnant’s outflow anticyclone with the midlatitude flow.
The injection of diabatically warmed upper-tropospheric air into the extratropics can lead to block (Rex 1950a, b) development, a flow regime that can have a deleterious effect on the skill of numerical guidance (Legras and Ghil 1985; Tibaldi and Molteni 1988; Tracton 1990). The diabatic generation of low potential vorticity (PV) at upper levels, effectively produced by convective outflow, has been shown to excite diabatic Rossby waves (Matthews and Kiladis 2000; Wernli 2004; Chen 2005) whose characteristics are particularly problematic for numerical models in which moist convection—and by extension the bulk upscale effects thereof—is poorly represented. The persistence and amplitude (Chen 2005) of these waves suggest that they can perturb the hemispheric flow on time scales of weeks to a month. It is shown here that Katrina’s outflow results in a blocking anomaly that not only modifies the Northern Hemispheric flow for 2 weeks, but also plays a role in the tropical cyclogenesis of Hurricane Nate (2005) and intensification of Hurricane Maria (2005) through the development of a low-latitude trough (Habjan and Holland 1995; Ferreira and Schubert 1999).
The global nature of the upper-tropospheric warm pool is evinced once its source regions are identified. Katrina’s outflow does not itself become ensconced in the midlatitude flow but rather plays an integral role in the evolution of the preexisting warm pool’s life cycle by establishing an anomalous connection between the Tropics and extratropics (referred to here as a conduit or freeway). The source of the warm pool itself is found to be a merger of the outflow from eastern Pacific Hurricane Hilary (2005) and a breakaway feature ejected from the south Asian anticyclone (SAA; Mason and Anderson 1963; Reiter and Gao 1982).
This paper begins with a description of the datasets used in the study in section 2. Section 3 summarizes the final stages of Hurricane Katrina’s life cycle. The influence of the warm pool on the flow over the North Atlantic Ocean, predictability, and the genesis and intensification of hurricanes Maria and Nate is investigated in section 4. Section 5 explores the origins of the upper-tropospheric warmth. The study concludes with a discussion of the findings in section 6. Table 1 lists the symbols used in the text and figures.
2. Datasets and methods
The track and intensity values for the hurricanes described in this study (Atlantic Hurricanes Katrina, Maria, and Nate, and eastern Pacific Hurricane Hilary) are based on advisory locations and strengths distributed by the National Hurricane Center (NHC). Analyses are performed using 0.5° National Centers for Environmental Prediction (NCEP) Global Forecast System (GFS) analysis data. This dataset is also used at 6-hourly intervals to compute forward and backward trajectories, with linear temporal interpolation used for 1-hourly substeps between analysis periods to enhance the accuracy of the trajectory calculation. The use of satellite brightness temperature in this study follows the methodology of Part I. Ensemble forecast data analyzed in section 4a is from the NCEP Medium Range Forecast ensemble set, archived with a 2.5° grid spacing.
Anomalies in this study are defined as departures from the 1968–96 long-term mean values, as noted in Part I. Potential vorticity is computed as
where ζθ is the relative vorticity on an isentropic surface, f is the local planetary vorticity, θ is the potential temperature, and p is the vertical pressure coordinate. The dynamic tropopause (DT) is once again defined as the 2.0-PVU (1 PVU ≡ 10−6 m2 K kg−1 s−1) surface and quantities with “DT” subscripts refer to the relevant variables at this level (Morgan and Nielsen-Gammon 1998). The diagnostic PV inversion undertaken in section 4 follows the piecewise method developed by Davis and Emanuel (1991), which uses Ertel (1942) PV and the nonlinear balance equation (Charney 1955) to relate anomalous PV to its corresponding streamfunction and geopotential fields.
A diagnostic based on the Eady model (Eady 1949) is presented in section 4 in which the model’s basic state is used as a dynamically significant background, preferable to climatology, from which to compute potential temperature anomalies at the upper boundary of the Eady model, here described by the dynamic tropopause (McTaggart-Cowan et al. 2006).
The anomaly correlation (AC) score is a measure of a deterministic model’s success in predicting deviations from climatological norms. In this study, the computation of AC is based on the method of Brankovic et al. (1990) and is computed as
wherein the metric in both the forecast and the analysis is unbiased with respect to climatology (taken to be the NCEP reanalysis long-term means; Kalnay et al. 1996). Here, M represents the chosen metric (usually 500-hPa heights) and σ is the standard deviation. Primes denote departures from climatology, and subscripts indicate whether the term refers to the forecast or verifying analysis. The AC scores can vary from a perfect 1 to a negatively “perfect” −1, with a useful range from 0.6 to 1 (Kalnay et al. 1991).
3. Hurricane Katrina’s landfall and ET
Despite the emphasis of this study on the nature of a long-lived warm pool, a brief summary of the latter stages of Hurricane Katrina’s life cycle is necessary. For a complete analysis, the reader is referred to Part I of this study.
Weakening from its peak intensity in the Gulf of Mexico, Katrina makes its third and final landfall near the Louisiana–Mississippi border at about 1445 UTC 29 August as a category-3 (Simpson 1974) storm (observed winds near 54 m s−1). The hurricane’s winds and storm surge cause severe damage to the city of New Orleans and to a broad swath of southern Louisiana and Mississippi, with hurricane-force winds reported as far as 170 km inland. The NHC downgrades Katrina to a tropical storm on 30 August as the system accelerates northeastward ahead of an approaching upper-level trough.
The ET of the weakening system on 30 August is very rapid, as strong upper-level winds between Katrina’s outflow anticyclone and the upstream trough shear the potentially warm air off the top of the system (Fig. 2). The remnant cyclonic vortex fails to reintensify following ET and eventually dissipates near the southern tip of Greenland on 7 September. However, Katrina’s diabatically generated anticyclone is advected eastward toward the western Atlantic Ocean on 30 August and 3 September before it tracks southward back into the Tropics. The circulation in this trough–ridge couplet, formed by the outflow anticyclone and the upstream trough that it helps to intensify, results in an anomalously strong southerly wind component across eastern North America over this period (Fig. 3). The effect of this conduit/freeway is to establish a connection between the Tropics and midlatitudes by advecting a transient warm pool (TWP) from its 1 September position over the southwestern United States up the eastern seaboard and into the extratropics. It is this legacy of Hurricane Katrina’s life cycle that will be investigated in this study.
4. Effects of the transient warm pool
The anomalous southwesterly upper-level flow over the eastern United States, established through the interaction between Katrina’s outflow anticyclone and the upstream midlatitude trough (Fig. 3 and feature “T” in Fig. 4), advects the TWP toward the western North Atlantic ocean by 0000 UTC 4 September (Fig. 4a). Over the next 4 days, the TWP evolves from a northeast–southwest-oriented ridge to an isolated warm pool on the dynamic tropopause as it progresses across the North Atlantic (labeled “W” in Fig. 4). The structure of the TWP is shown in Fig. 5 and is characterized by a strong potential temperature anomaly at upper levels (Fig. 5a) compared to the Eady model (Eady 1949) basic state (outlined in section 2) and a deep warm anomaly (relative to both climatology and the surrounding atmosphere) extending throughout the column beneath the elevated tropopause (Fig. 5b). The reversal of the gradient of θDT satisfies the Pelly and Hoskins (2003) criterion for a blocking feature as shown in Fig. 6 between 0000 UTC 3 September and 0000 UTC 7 September, indicating a significant blocking event in the region. The TWP, and its associated blocking, has two important ramifications for the flow over the North Atlantic that will be described in this section: a degradation of forecast skill (section 4a) and the genesis and development of hurricanes Nate and Maria (labeled “M” in Fig. 4), respectively (section 4b).
The life cycle of the TWP does not end following the breakdown of the block over the North Atlantic. The feature tracks northeastward over the Baltic Sea on 8 September, and from there progresses into Russia as a coherent upper-level disturbance (Fig. 7). Between 9 and 12 September, the TWP moves across western Russia and into Siberia, before finally becoming reintegrated into the tropical high-θDT source shortly after 0000 UTC 12 September over the Tibetan Plateau. It will be shown in section 5b that this region is climatologically favorable for the development of high θDT values because of the effects of insolation on the elevated, low albedo surface of the plateau.
The effect of atmospheric blocking on predictability has been shown by Legras and Ghil (1985), Tibaldi and Molteni (1988), and Tracton (1990), among others. In particular, the “breaks”—regime transitions—between blocking and zonal patterns are difficult for numerical weather prediction models to accurately assess (Namias 1982), especially if the causes for the break relate to internal dynamics rather than to external forcings. The interaction of the larger-scale (planetary) slowly growing modes that characterize the block itself, with the smaller-scale more rapidly growing modes in the flanking troughs make the prediction of future states beyond the short-range problematic. These nonlinear wave–wave interactions (Egger 1978) between features of planetary and “medium” scales can be responsible for both the maintenance and the destruction of the blocking regime (Tracton 1990). Errors in the representation of medium-scale waves (defined as total wavenumbers 7–12) can propagate upscale over the short-range forecast period so as to adversely affect the planetary-scale background in medium-range predictions. Tracton (1990) finds this to be especially true during a blocking regime transition in a full numerical weather prediction model, although the influence of the errors can be reduced with increased horizontal resolution.
The use of ensemble spread as an indicator of predictability (the spread–skill relationship; Kalnay and Dalcher 1987; Buizza 1995, 1997) has been shown to be more valid for medium-range predictions than for short-range forecasts (Stensrud et al. 1999). For this reason, the ensemble forecasts investigated in this study are initialized at 0000 UTC 1 September, 48 h before the onset of blocking and 144 h before the break in the blocking pattern (Fig. 6), placing the bulk of the forecast period of interest in the medium range. The ensemble mean 48-h θDT forecast shows the TWP located over the Canadian Maritimes at 0000 UTC 3 September (Figs. 8a,b), with the highest spread values associated with the location and intensity of the trough features over Quebec and the central North Atlantic. By 0000 UTC 4 September (Figs. 8c,d), the ensemble mean TWP is separated from its tropical θDT source and has begun to effectively block the North Atlantic flow, as evinced by the reversal in the meridional θDT gradient. The primary uncertainty at this time is associated with the downstream cyclone centered near 15°W; however, a band of enhanced spread (1–1.5 normalized standard deviations from the ensemble mean) encompasses the periphery of the TWP. By 0000 UTC 5 September, this region of reduced predictability has grown in both strength and area, surrounding the TWP with ensemble spreads that are up to three normalized standard deviations from the ensemble mean for the 96-h forecast (Figs. 8e,f). The 120-h forecast (Figs. 8g,h, verifying at 0000 UTC 6 September) shows a weakened ensemble mean TWP feature whose zonal displacement appears to be a primary cause of the ensemble spread. Clearly, the injection of the TWP into the midlatitude flow at 0000 UTC 2 September has a severe impact on medium-range predictability over the North Atlantic.
The rapid increase in ensemble spread at the periphery of the TWP, rather than in the interior of the anticyclonic feature, is a result of both the ensemble initialization scheme and the locally large gradients in the analysis metric (θDT). The breeding method used at NCEP (Toth and Kalnay 1997) focuses initial condition perturbations on the fastest-growing modes in the individual ensemble members during the optimization period. Because these modes tend to be baroclinic in nature, they are commonly associated with trough and frontal features. The growth of perturbations introduced in the ridge environment of the TWP would quickly be outpaced by that of the baroclinic modes at the interface between the TWP and its flanking troughs. This is exactly what occurs in Fig. 8 as the uncertainty associated with the TWP is manifested as increased ensemble spreads at its lateral boundaries. Large gradients in θDT will tend to amplify peripheral ensemble spread for a given displacement of the TWP; however, the following analysis of deterministic GFS forecasts shows that the reduction in forecast skill is a robust result even for the spatially smooth 500-hPa height field.
The reduction in predictability over the period of TWP passage and blocking across the North Atlantic Ocean can be shown deterministically using the AC diagnostic described in section 2. The mean and standard deviation of the GFS AC scores over the domain bounded by 20°N, 80°W and 80°N, 20°E (shown with a heavy box in the inset of Fig. 9) are presented in Table 2. The forecast is generally considered useful until the AC score for the region of interest falls below 0.6 (Kalnay et al. 1991), day 6 for the August–October 2005 period over the North Atlantic (Table 2). Figure 9 shows the deviations of the AC scores from these mean values for the 2.5-week period covering the TWP’s passage across the North Atlantic region. Above-average deterministic predictability prevails for the week leading up to the TWP blocking, followed by a sudden decrease in skill during the TWP’s passage. Although the recovery of predictability following the blocking event is not as sharp as its reduction during the onset, AC values return to near normal following the exit of the TWP from the domain. As noted by Namias (1982), the maximum reductions in predictive skill appear to focus on the breaks between blocked and zonal flow regimes. These results suggest that the maxima in ensemble spread shown in Fig. 8 are indicative of true predictability reduction rather than reflective of locally large gradients in the analysis metric.
b. Tropical cyclogenesis
In addition to the TWP’s influence on predictability over the North Atlantic ocean, the broad upper-tropospheric anticyclone has a strong effect on the dynamics and thermodynamics of the region. As the TWP enters the midlatitudes at 0000 UTC 2 September—driven by the anomalous southwesterly flow that is the legacy of Hurricane Katrina’s outflow (labeled “O” in Fig. 10) following ET, as described in section 3 and shown in Fig. 3—its influence on the upper-level flow over the North Atlantic is evident (Figs. 10c–f). A broad region of northerly flow of up to 25 m s−1 exists along the eastward edge of the TWP (labeled “W” in Fig. 10) as the wind circulates anticyclonically around this warm upper-level feature. This northerly flow drives a band of lowered dynamic tropopause (350 hPa) southward and westward into the Tropics at 55°W and separates it from its PV tail (identified with a dotted line in Fig. 10). The effect of the intrusion of positive PV anomalies of extratropical origin on tropical cyclogenesis and tropical transition has recently been studied by Bracken and Bosart (2000), Davis and Bosart (2003, 2004), among others. Historically, these features were known as tropical upper-tropospheric troughs (TUTTs; Sadler 1976, 1978). Ferreira and Schubert (1999) show that TUTT cells can develop downshear of TCs entering the midlatitudes through both the dispersion of short Rossby wave energy and the direct advective effects of the outflow anticyclone (Habjan and Holland 1995); the latter TUTT development mechanism is investigated in detail here with respect to the TWP.
The advection into the Tropics of the small-scale upper-tropospheric cold trough that characterizes the fragmented PV tail in Fig. 10 has a strong impact on the local environment. Specifically, the southwestward relocation of the PV tail by the TWP reduces the convective stability of the area by cooling the upper levels, provides a local cyclonic circulation in which vorticity can be readily generated by stretching due to convection, and enhances the coupling between the upper levels and the warm surface.1 As the upper-level cold trough interacts with its tropical environment from 3 to 5 September—most notably an easterly wave that crossed the African coast on 30 August (feature “N” in Fig. 10f)—the NHC identifies Tropical Depression 15 at 1800 UTC 5 September near 28°N, 68°W. The incipient tropical cyclone N becomes Tropical Storm Nate 6 h later (0000 UTC 6 September; Fig. 11). As was the case for Hurricane Katrina itself (see Part I for details), it appears as though Hurricane Nate had both tropical (easterly wave) and extratropical (PV tail) features involved in its cyclogenetic process.
To isolate the effect of the TWP on the tropical forcings that resulted in the development of Nate, the flow solely attributable to the warm upper-tropospheric feature is isolated for a diagnosis of Hurricane Nate’s genesis. The piecewise PV inversion technique of Davis and Emanuel (1991) is employed to separate the TWP’s flow from the full field. The source domain for the PV anomaly to be inverted is shown with the heavy outline in Fig. 10f and encompasses the entire TWP at 0000 UTC 3 September over the eastern United States. All levels at or above 500 hPa are used in the inversion procedure, thereby capturing the full depth of the warm pool. Only the anomalous anticyclonic PV is inverted, representing the majority of the subdomain given the extent of the TWP anomaly.
The structure of the lower-level easterly wave system at 0000 UTC 3 September is shown in Fig. 12a (labeled “E”) and is characterized by a north–south-elongated center of elevated relative vorticity with a maximum at its northern tip. Figure 12b shows the component of the flow on the dynamic tropopause uniquely attributable to the TWP. Westerly advection of the PV tail remnant is evident above the slowly propagating lower-level feature E (Fig. 12a). The effect of this advection can be described using quasigeostrophic (QG) dynamics. Although QG theory is incapable of describing much of the tropical atmosphere, it offers the advantage of implicitly separating tropical from traditional “midlatitude” ascent forcings and has been used extensively in studies of baroclinic precursors to tropical cyclogenesis studies (Bosart and Bartlo 1991; Bracken and Bosart 2000; McTaggart-Cowan et al. 2006). It is precisely the QG forcings that are fundamental to the role of the tropopause depression in this genesis event. Accordingly, the diabatic terms of the QG omega equation are dropped to yield
Here, ω is the pressure coordinate vertical motion, σ ≡ −(1/ρθ)(∂θ/∂p) is the static stability (θ is potential temperature and ρ is dry air density), f0 is a reference Coriolis parameter ( f0 = 10−4 s−1 for the TWP), R is the gas constant for dry air, υg is the geostrophic wind vector, ζg is the geostrophic vorticity, and T is the dry air temperature. All derivatives are computed on pressure surfaces, as noted with a subscript p.
The upper-level QG ascent forcing [the two terms on the right-hand side of Eq. (3) evaluated between 400 and 100 hPa] at 0000 UTC 3 September is shown in Fig. 12a, with positive values indicative of areas preferred for ascent. A maximum in QG forcing exists near 70°W in a broad northeast–southwest-oriented swath associated with the advection of a weak upper-level trough. Embedded within this forcing region is a local maximum that is centered over the northern component of the easterly wave feature E (Fig. 12a). Although the vertical motion implied by this forcing is too weak to initiate tropical cyclogenesis by itself, the superposition of this TWP-induced ascent and the easterly wave results in the development of Hurricane Nate.
The other feature of interest in Fig. 12a is Tropical Storm Maria (denoted with a tropical storm symbol). Although Maria develops into a tropical depression on 1 September (Fig. 11), it strengthens slowly until reaching hurricane strength at 0600 UTC 4 September. Initially beneath unfavorable TWP-induced ascent forcing (not shown), Maria is surrounded by upward QG forcing by 0000 UTC 3 September (Fig. 12a). The influence of the TWP on the flow over the North Atlantic is highlighted by the fact that the only two regions in which the TWP-induced ascent forcing is superposed on easterly wave features are the two areas in which tropical cyclogenesis is under way.
5. Origins of the TWP
The TWP, injected into the midlatitudes by the anomalous upper-level conduit/freeway established by the ET of Hurricane Katrina (Fig. 3), has been shown to exert a strong forcing on the flow throughout the midlatitude component of its life cycle. It reduces medium-range predictability over the North Atlantic (Fig. 8), plays a role in the genesis and development of hurricanes Nate and Maria (Fig. 12), respectively, and follows a 10-day track across northern Europe and Asia (Fig. 7). The origin of such an important feature clearly warrants further investigation. In this section, the TWP is found to arise from the merger of a pair of diabatically generated warm pools, one created by eastern Pacific Hurricane Hilary and the other by the SAA, the same feature into which the TWP is eventually absorbed on 12 September. This merger of the source components takes place over the southern United States, resulting in the TWP’s placement in the entry region of the southwesterly conduit/freeway that begins its journey into the midlatitudes in early September.
At lower latitudes, the components of the TWP are difficult to track subjectively because of the elevated θDT values throughout the equatorial regions. As a result, most of the analyses in this section rely initially on back trajectories of representative parcels within the TWP to determine the source region for the feature. Once the trajectories have highlighted an area of potential importance, the dynamical linkages between the source region and the final warm anomaly can be determined in order to support the findings of the trajectory-based diagnostic. Figure 13 shows a set of 7-day back trajectories for the TWP verifying at 300 hPa at 0000 UTC 2 September. The core of the TWP anomaly is composed of air parcels from two distinct source regions. The first, enclosed by the box over the eastern North Pacific in Fig. 13, is associated with a strong ridge that builds in situ over the area as a result of the outflow from Hurricane Hilary. The second source region, contained at the southern end of the box over the central North Pacific in Fig. 13, is associated with a warm pool ejected from the SAA and is less well defined at 0000 UTC 26 August.
a. Hurricane Hilary: Contribution to the TWP
Hurricane Hilary forms near 13°N, 95°W, approximately 600 km west of El Salvador at 1200 UTC 19 August (Hilary’s track is shown by the heavy magenta line in Fig. 14a). Moving slowly northwestward parallel to the Mexican coastline, Hilary did not intensify to category-1 strength until 21 August. Although tropical storm–force winds are experienced along portions of the west coast of Mexico as Hilary reaches its peak intensity of 45 m s−1 (category 2) at 0000 UTC 22 August, the hurricane continues on a northwesterly track and moves slowly away from the coast. Weakening begins on 23 August, and Hilary once again becomes a tropical storm at 1800 UTC 24 August. This slow weakening trend continues and deep convection is still present near the storm’s center until 26 September, when the lower-level circulation is reduced to a convectively inactive depression.
The effect of Hurricane Hilary on upper-tropospheric warmth is shown in Fig. 14. At 0000 UTC 25 August, just after Hilary has been downgraded to a tropical storm, there is still widespread convection to the west of the surface center (indicated with the tropical storm symbol in Fig. 14). The anticyclonic curvature of the flow on the dynamic tropopause to the north and west of Hilary suggests that the warm outflow from the storm is warming tropopause potential temperatures as far north as 35°N. Ridge building continues as the remnant circulation progresses northwestward, as reflected in the evolution of the θDT ridge that erupts west of California between 0000 UTC 25 August (Fig. 14a) and 0000 UTC 28 August (Fig. 14b). The nonconservative nature of this structural change of the θDT field suggests that most of the forcing for this ridging arises from diabatic processes, particularly convection associated with the decaying tropical system. Hilary’s ridge progresses slowly eastward between 0000 UTC 28 August and 0000 UTC 2 September and merges with the SAA warm pool component to create the TWP (Fig. 13).
b. South Asian anticyclone: Contribution to the TWP
The origin of the warm pool in the western source region of Fig. 13 is less obvious than that associated with Hilary’s outflow. Although the trajectories in Fig. 13 begin just to the north of the main tropical warm source on the dynamic tropopause (confined equatorward of 10°N in Fig. 13) near the date line, the coherency of the warm pool feature is difficult to discern because of the high θDT values of its surrounding environment. However, analysis of a series of dynamic tropopause plots (Fig. 15) beginning at 0000 UTC 17 August, shows that a warm pool initially centered near 45°N and 120°E (labeled as a TWP component “W” in Fig. 15) breaks off the SAA (labeled “S” in Fig. 15) and moves slowly eastward before splitting to flow around a compact extratropical cyclone centered at 40°N, 170°E at 0000 UTC 19 August (Fig. 15f). One branch of the warm pool dives southward toward the Tropics, while the other is advected northward downshear of the cyclone. As shown in Fig. 16 it is primarily the southern branch of the W warm pool that contributes to the western source region of the TWP (bounded by the heavily bordered box in Fig. 16). Although some parcels in the northern branch curl anticyclonically to reenter the TWP source region, many of them continue northeastward to form an upper-level anticyclone near the west coast of North America by 0000 UTC 26 August (Fig. 16).
The ultimate source of the warm pool in the western source region for the TWP is therefore the expulsion of air with elevated θDT values from the SAA, a persistent feature that has received relatively little attention in the recent western literature. However, early studies by Mason and Anderson (1963), Flohn (1965), and Reiter and Gao (1982), among others, explore the structure and evolution of the SAA and its relationship with the south Asian monsoon. The unique nature of the Tibetan Plateau, with a mean elevation of 5400 m, yields surface pressures on the order of 600 hPa. This, coupled with sparse vegetation and a generally low albedo surface, ensures that the effects of insolation—and the resulting sensible and latent heat fluxes from the surface—will be of primary importance to the local atmosphere. Reiter and Gao (1982) show in a pair of case studies that heat lows at the surface (caused by either direct flux-driven warming or downsloping on the northern side of the Himalayas) induce warm-core vortices based in the planetary boundary layer whose upper-level (300 hPa and above) outflows are enhanced by diabatic releases of latent heat. The SAA develops in April and is associated with the breakdown of the subtropical branch of the westerly jet to the south of the Himalayas that acts as a precursor to the onset of the monsoon flow in the area. Mason and Anderson (1963) find that the SAA reaches its peak intensity in July and August, consistent with theories of the heat flux-driven maintenance of the circulation.
Longitudinal displacements of the SAA occur frequently throughout the year despite the anticyclone’s dominance as the primary wavenumber-1 feature in the Northern Hemisphere summer (Mason and Anderson 1963). The SAA periodically meanders west as far as the Caspian Sea (50°E) and east as far as the western edge of the Tibetan Plateau (120°E). Mason and Anderson (1963) show that the breakdown of the SAA in 1957 (one of the two years studied) is associated with the eastward ejection of the warm pool at the end of August, and its replacement by a weaker anticyclone in early September. This fracturing and eastward propagation of the SAA feature during its breakdown process at the end of August appears to characterize the generation and evolution of the high θDT values that form the western source component of the TWP.
The maintenance of the TWP against the effects of radiation that cool upper-level parcels at approximately 1–2 K day−1 must be accomplished by a nonconservative process. As shown in a time series of parcel potential temperatures (Fig. 17a), the basic cooling rate is replaced by sustained heating of approximately 5 K day−1 for a subset of the parcels, primarily those that comprise the TWP. As shown in Fig. 17b, both this warming and the change in atmospheric structure along the trajectory of a representative parcel are consistent with moist convection–induced heating in the column. The capped, conditionally unstable profile at 0000 UTC 22 August displays a subsidence inversion above 800 hPa and a stability tropopause near 200 hPa. By 0000 UTC 25 August, the TWP sounding is almost moist neutral, with the stability tropopause elevated to 150 hPa, indicative of the presence of moist convection. Further evidence for deep convection is provided by infrared satellite imagery for 1200 UTC 24 August (Fig. 17c), which shows a broad area of high cloud tops collocated with the rapidly warming TWP parcels. This episodic reinvigoration of the TWP by moist convection is crucial to the maintenance of the feature against the effects of radiative cooling over its extended lifetime.
6. Summary and discussion
The life cycle of Hurricane Katrina (2005) includes tropical cyclogenesis, preliminary landfall followed by little weakening over land, rapid intensification over the Gulf of Mexico, devastating landfall near New Orleans, and nonreintensifying ET. While all attention is focused on the Louisiana coast following the storm’s passage, the innocuous ET of the system leads to an anomalous connection between the Tropics and the midlatitudes as the upper-level outflow anticyclone interacts with an approaching baroclinic trough. In the entry region of this conduit/freeway, a preexisting upper-tropospheric warm anomaly (TWP) acquires a northerly track component and begins to accelerate into the westerlies. The motivating question for this research is therefore, “can an ET event that does not involve reintensification exert a significant influence on the midlatitude flow?” The affirmative answer has a significant impact on forecasting since the effects of ET on the extratropical flow are generally considered to be confined to the reintensification of the system itself and the potential for downstream development.
The life cycle of the TWP is summarized in Fig. 18 from 14 August to 11 September. The TWP enters the midlatitudes on 3 September and forms a blocking structure over the North Atlantic Ocean. As shown in Fig. 8, this pattern proves problematic for medium-range predictability based on analysis of the ensemble spread–skill relationship. The anticyclonic flow around the TWP as it moves off the North American east coast results in strong northerly flow over the western half of the Atlantic Ocean that extends into the Tropics (small blue arrows in Fig. 18). The interaction of this flow with a preexisting PV tail that stretches across the equatorial Atlantic (stippled black line in Fig. 18) results in regions of enhanced QG ascent forcings that are collocated with lower-level easterly wave features and appears to aid in the development of Hurricane Nate and the intensification of Hurricane Maria (purple and red dots in Fig. 18, respectively).
The origin of the TWP is traced to two independent source regions in the tropical Pacific: the diabatically generated warm outflow from Hurricane Hilary and a warm pool near the date line. The latter is shown to have originated as an extrusion from the SAA during the preliminary late-August breakdown of the diabatically maintained anticyclonic circulation over the Tibetan Plateau (Fig. 18). The combination of these two warm anomalies results in the formation of the TWP over the southwestern United States in early September, just as Hurricane Katrina is undergoing ET and its outflow is working with the upstream trough to establish a southwesterly conduit/freeway over eastern North America. The fortuitous location of the TWP in the entrance region of this anomalous upper-level flow results in its acceleration into the midlatitudes.
The hemispheric scale of the features involved with the life cycle of the TWP suggests that the details of its evolution are unique. However, its influence on tropical cyclogenesis and predictability once it enters the midlatitudes is important and an understanding of the processes involved may enhance the forecasting of broadly similar events in the future. The answer to the question of whether a nonreintensifying ET event can exert a significant influence on the midlatitude flow is an emphatic “yes,” with the caveat that it is not the remnant outflow of Hurricane Katrina itself that comprises the long-lived TWP. Instead, the crucial role played by the ET of Katrina is that of establishing the anomalous connection between the Tropics and extratropics that permits the injection of the TWP into midlatitudes. It is therefore important from a forecasting perspective to maintain a regional situational awareness rather than focusing entirely on the possible reintensification of the ET event. Even seemingly transient modification of the flow by a decaying ET can result in persistent adjustment of the large-scale pattern under favorable conditions.
The authors thank Robert Hart for providing the CPS figure for Hurricane Katrina, and John Molinari and Tim Hewson for helpful discussions during the preparation of the manuscript. This work was supported by NSF Grant ATM0304254, the Canadian Foundation for Climate and Atmospheric Sciences, and the Canadian Natural Sciences and Engineering Research Council.
Corresponding author address: Ron McTaggart-Cowan, University at Albany, State University of New York, DEAS-ES351, Albany, NY 12222. Email: email@example.com