Tropical Transition of an Unnamed, High-Latitude, Tropical Cyclone in the Eastern North Pacific 1. Introduction Tropical cyclones (TCs) are not exclusive to the tropics. While the environmental conditions deemed favorable for tropical cyclogenesis by Gray (1968) and DeMaria et al. (2001) are typical observed at tropical latitudes, environmental conditions can become favorable for tropical cyclogenesis in locations removed from the tropics. The global climatology of tropical cyclogenesis events constructed by McTaggartCowan et al. (2013) reveals that many TCs forming poleward of 30°N (25°S) in the Northern (Southern) Hemisphere during 1948–2010 developed in the presence of an upper-tropospheric disturbance in a baroclinic environment (their Fig. 7). These cases of baroclinically induced tropical cyclogenesis are typically associated with the tropical transition (TT) process (Davis and Bosart 2003, 2004), during which an extratropical cyclone (EC) transitions into a TC. In the initial stages of the TT process, vertical wind shear in a baroclinic environment produces a region of upward motion that focuses deep convection and diabatic heating. Vertical wind shear values are subsequently reduced by the diabatic redistribution of potential vorticity (PV) in the vertical (Raymond 1992) and by divergent outflow in the upper troposphere, allowing the surface cyclone to intensify via wind-induced surface heat exchange (Emanuel 1986, 1995). TCs forming via the TT process have been documented in many of the ocean basins discussed by McTaggart-Cowan et al. (2013), including the western North Atlantic (e.g., Moore and Davis 1951; Bosart and Bartlo 1991; Bracken and Bosart 2000; Davis and Bosart 2003, 2004; McTaggart-Cowan et al. 2006; Guishard et al. 2007; Evans and Guishard 2009; Guishard et al. 2009; Hulme and Martin 2009a,b), western South Atlantic (e.g., Pezza and Simmonds 2005; McTaggart-Cowan et al. 2006; Evans and Braun 2012), and western South Pacific (e.g., Garde et al. 2010). 1 TCs forming via the TT process have also been documented over large bodies of water that are not explicitly discussed by McTaggart-Cowan et al. (2013), including the Great Lakes (e.g., Sousounis et al. 2001) and Mediterranean Sea (e.g., Ernst and Matson 1983; Pytharoulis et al. 1999; Reale and Atlas 2001; McTaggart-Cowan et al. 2010). Tropical cyclogenesis events occurring over the western South Atlantic, Great Lakes, and Mediterranean Sea are extremely rare. The infrequent development of TCs in these regions is likely associated with the presence of relatively cold sea surface temperatures (SSTs) that do not exceed the 26.5°C threshold for tropical cyclogenesis identified by Gray (1968). In order to facilitate the development of deep convection necessary for an EC to undergo TT, tropospheric lapse rates must steepen in response to upper-tropospheric cooling associated with an encroaching upper-tropospheric disturbance (Davis and Bosart 2003, 2004). In late October 2006, an unnamed TC (hereafter Invest 91C) developed at ~40°N over the eastern North Pacific. A weak EC, forming downstream of a thinning uppertropospheric trough over the Gulf of Alaska, served as the precursor disturbance that would ultimately undergo TT. The TT of Invest 91C, which took place between 0000 UTC 29 October 2006 and 0000 UTC 2 November 2006, was extremely unusual—occurring over ~16°C SSTs in a region historically devoid of TC activity (Fig. 1). A synoptic overview of the formation of Invest 91C is presented in the following section to document the features and processes associated with its development and TT. Model simulations of the TT of Invest 91C will also be presented to explore how the use of different microphysical parameterization schemes could affect the structure and intensity of an EC undergoing TT within a numerical model. Information on model configuration, as well as the data and methodology used to construct the model simulations, will be presented in section 3. Section 4 will discuss specific findings from the model simulations, including the observed differences between model runs. 2 This paper will conclude with a brief discussion and presentation of ideas for future research. 2. Synoptic overview http://www.atmos.albany.edu/student/abentley/research_images/ne_pac_tc/cfsr_pacific.html The upper-tropospheric flow pattern over the central North Pacific becomes highly amplified in late October 2006 (hereafter, all dates are in 2006) in response to two ECs that develop over eastern Asia. An ~1012-hPa EC (EC1) begins to deepen off the southeastern coast of Russia at 0000 UTC 25 October, downstream of a progressive upper-tropospheric PV anomaly. At the same time, an ~1008-hPa EC (EC2) forms along the east coast of Japan on the southwestern edge of the remnants of a midlatitude cold front—beneath the fracturing equatorward end of an upper-tropospheric trough. EC1 moves ~1500 km to the east-northeast over the following 48 h, deepening to ~1004 hPa over the Kamchatka Peninsula. EC2 begins to approach EC1 during this period, moving ~3000 km to the northeast and deepening by ~8 hPa. The poleward advection of high potential temperatures on the dynamic tropopause (DT) downstream of EC1 and EC2 aids in the formation and amplification of an upper-tropospheric ridge over the central North Pacific between 0000 UTC 25 October and 0000 UTC 27 October. Enhanced northwesterly flow downstream of the amplifying ridge aids in the formation and amplification of an upper-tropospheric trough to the south of the Aleutian Islands during this period. http://www.atmos.albany.edu/student/abentley/research_images/ne_pac_tc/cfsr_alaska.html EC1 and EC2 merge together by 0000 UTC 28 October, forming a sub-992-hPa EC (EC3) over the Bering Sea. Persistent northwesterly flow on the eastern periphery of the central North Pacific ridge aids in the stretching and thinning of the upper-tropospheric trough in the southern Gulf of Alaska. Negative PV advection by the 300–200-hPa layer-averaged irrotational wind occurs to the west and east of the upper-tropospheric trough at 0000 3 UTC 28 October (Fig. __), tightening the horizontal PV gradient and transforming the upper-tropospheric trough into an upper-tropospheric PV streamer. The surface cyclone that will ultimately become Invest 91C begins to develop along the southeastern edge of the upper-tropospheric PV streamer by 1200 UTC 28 October, in the equatorward entrance region of a 250-hPa jet streak. Divergent outflow over the center of the surface cyclone, indicated by the starburst pattern in the 300–200-hPa layer-averaged irrotational wind field emanating from a region of 600–400-hPa ascent, opposes the eastward progression of the southern portion of the upper-tropospheric PV streamer. The southern portion of the upper-tropospheric PV streamer fractures from the northern portion by 0000 UTC 29 October. The surface cyclone, now positioned slightly to the northeast of the uppertropospheric PV anomaly, deepens to ~1000 hPa between 1200 UTC 28 October and 0000 UTC 29 October. Warm lower-tropospheric air, manifested as high 1000–500-hPa thickness values, wraps around the east side of the surface cyclone, reversing the meridional temperature gradient and producing a bent-back warm frontal structure on the northwestern periphery of the surface cyclone by 1200 UTC 29 October. As previously documented by Hulme and Martin (2009b), the bent-back warm front plays an important role in the TT of an EC. Convection along the bent-back warm front is believed to generate lower-tropospheric vorticity on the western half of the cyclone. This enhanced lower-tropospheric vorticity intensifies cold air advection on the northern and western sides of the cyclone and helps to isolate the cyclone’s developing warm core. The diabatic redistribution of PV in the vertical along the bent-back warm front, upshear from the center of the cyclone, also helps to reduce vertical wind shear values over the center of circulation. The coupling index, defined in Bosart and Lackmann (1995) as the difference between the potential temperature of the DT and equivalent potential temperature at 850 hPa, is shown in Fig. ___. Figure ___ reveals extremely low values of the coupling index (< −5 K) near the center of the surface cyclone at 1200 UTC 29 October, indicating the presence of highly unstable air in the midtroposphere. 4 A vertical cross section, taken through the center of the surface cyclone at 1200 UTC 29 October, emphasizes the upper-tropospheric contribution to the midtropospheric instability. An upper-tropospheric PV anomaly, associated with the southern portion of the upper-tropospheric PV streamer, extends below 500-hPa just to the south of the center of the surface cyclone. Potential temperature contours beneath the upper-tropospheric PV anomaly are widely spaced and bow upward, indicating (1) the instability of the midtroposphere near the center of the surface cyclone depicted in Fig. ___ and (2) that the cyclone is primarily cold core. The upper-tropospheric PV anomaly becomes collocated with the center of the surface cyclone between 1200 UTC 29 October and 0000 UTC 30 October. Weakly negative 925–500-hPa thermal vorticity values associated with the center of the cyclone have separated from the bent-back warm front by 0000 UTC 30 October, suggesting that the cyclone is losing its frontal structure and is beginning to acquire more TC-like characteristics. The region of 925–850-hPa cyclonic relative vorticity associated with the center of the cyclone breaks away from the warm-frontal band between 0000 UTC 30 October and 1200 UTC 30 October. The expansion of negative 925–500-hPa thermal vorticity values near the center of the cyclone and the reduction of positive 925–500-hPa thermal vorticity values in the surrounding area indicate that the cyclone is becoming less cold core. The region of 925–850-hPa cyclonic relative vorticity associated with the center of the cyclone remains collocated with the southern portion of the upper-tropospheric PV streamer and separate from remnants of the warm-frontal band between 1200 UTC 30 October and 1200 UTC 31 October. An upper-tropospheric trough approaches the transitioning cyclone from the central North Pacific during this period, wrapping around the southwestern edge of the storm. The central pressure of the transitioning cyclone falls below 996 hPa by 1200 UTC 31 October, indicating that the cyclone is deepening. 5 The cyclonic circulation associated with the upper-tropospheric trough approaching the transitioning cyclone from the central North Pacific is associated with the transitioning cyclone’s turn to the northwest between 1200 UTC 31 October and 1200 UTC 1 November. Thermal vorticity values surrounding the center of the transitioning cyclone are predominately negative by 1200 UTC 1 November, indicating that the storm has become warm core in the lower-to-midtroposphere. The transitioning cyclone’s lack of frontal structure and warm-core characteristics caused insert correct agency here to label the storm “Invest 91C” at 1200 UTC 1 November 2006. Invest 91C has completely transitioned into a sub-992-hPa, axisymmetric, warm-core TC by 0000 UTC 2 November (Figs. __a–d). GOES-10 visible satellite imagery, taken at approximately 0000 UTC 2 November, reveals the presence of an eye-like feature over the center of the cyclone (Fig. __). A vertical cross section, taken through the center of Invest 91C, reveals that the cyclone has completely undergone TT at this time. Potential temperature contours bowing down over the center of Invest 91C (Fig. __) confirm the warm-core structure of cyclone suggested by the 925–500-hPa thermal vorticity field (Fig. __). The upper-tropospheric PV anomaly that extended below 500 hPa at 1200 UTC 29 October (Fig. __) has been eroded by deep convection and no longer exists at 0000 UTC 2 November. A PV tower, also indicative of the diabatic redistribution of PV in the vertical, is present over the center of the cyclone between 925 hPa and 400 hPa. Equivalent potential temperature contours are vertically oriented on either side of the PV tower, suggesting that the eye-wall of the TC is well mixed and that deep convection has been occurring. Despite obtaining the characteristic structure of a TC, Invest 91C was not upgraded from an invest area to a TC during its life cycle. 6 Invest 91C weakens from ~992-hPa at 0000 UTC 2 November to ~1000 hPa at 1800 UTC 3 November, moving to the northeast at ~40 km h−1 and making landfall along the northwest coast of Washington. Figure ___ depicts surface data obtained from the Destruction Island, WA, buoy (DESW1), during the landfall of Invest 91C. DESW1 reported a ~6 hPa pressure drop between 1200 UTC and 1800 UTC 3 November as the center of Invest 91C passed. This ~6 hPa pressure drop coincides with a ~30 kt increase in sustained wind speeds measured by the buoy, with a maximum sustained wind speed of > 50 kts recorded at ~1600 UTC 3 November. 3. Model description and evaluation To better understand how the use of different microphysical schemes could affect the structure and intensity of an EC undergoing TT within a numerical model, a numerical simulation of the TT of Invest 91C is performed using version 3.4 of the Advanced Research Weather Research and Forecasting (WRF) modeling system (ARW; Skamarock et al. 2008). WRF simulations are initialized with 1° Global Forecast System final (FNL) analysis data beginning at 0000 UTC 28 October (correct start date? Or 24 h later? Up to you. See ATM611 paper images.) and ending at 0000 UTC 2 November. A two-way nested grid is used in this study with 30 km (10 km) horizontal resolution within the outer (inner) nest (Fig. __). Thirty-five vertical levels are analyzed. The WRF physics package allows the user to employ various combinations of cumulous, land surface, planetary boundary layer, and microphysical parameterization schemes. All WRF simulations performed in this study use the Kain-Fritsch cumulous parameterization scheme (Kain and Fritsch 1993), Noah land surface scheme (Ek et al. 2003), and and Mellor-Yamada-Janjic TKE planetary boundary layer scheme 7 (Mellor and Yamada 1982). This is done to isolate the impact of varying the microphysical parameterization (MP) scheme. The warming of the core of a TC is usually due to a combination of the diabatic heating in the eyewall and dry adiabatic descent within the eye. Stern and Nolan (2012) suggested that changes in the structure of the core of the TC may be sensitive to the distribution of diabatic heating and, therefore, the MP scheme used in a numerical model. Three simulations are performed in this study, each with a different MP scheme. Simulations are run with either the WRF Single-Moment 6-class (WSM6) (Hong and Lim 2006), WRF Single-Moment 3-class (WSM3) (Hong et al. 2004), or Kessler (Kessler 1969) MP scheme, consistent with the simulations performed by Stern and Nolan (2012). WSM6 has the highest complexity, predicting six categories of hydrometers: vapor, rain, snow, cloud ice, cloud water, and graupel. WSM3 is simplified, predicting three categories of hydrometers: vapor, cloud water/cloud ice, and rain water/snow. Melting (freezing) occurs instantaneously at temperatures above (below) freezing. WSM3 is often referred to as a “simply ice” scheme in which liquid and solid water cannot coexist. The Kessler MP scheme is a warm cloud scheme that includes vapor, cloud water, and rain. The only mechanism that produces new precipitation in the Kessler parameterization is the autoconversion process. Kessler is the simplest and most unrealistic of the three schemes considered in this study. 4. Analysis It is often necessary for WRF model output to be compared to surface observations and/or satellite data from the analyzed event to test the accuracy and validity of the solution. There is an unfortunate lack of radar coverage, upper air data, and surface observation stations over the portion of the eastern North Pacific where Invest 91C underwent TT. 8 For this reason, the output fields analyzed in this study will be compared to surface analyses from the National Centers for Environmental Prediction (NCEP) and archived GOES-11 infrared (IR) satellite imagery from the Cooperative Institute of Meteorological Satellite Studies’ (CIMSS) satellite blog, available online at http://cimss.ssec.wisc.edu/goes/blog/archives/211. Maximum reflectivity and mean sea level pressure values from the final output time of each model run (1200 UTC 1 November) are shown in Fig. __. The corresponding IR imagery is depicted in Fig. __. All three MP schemes correctly identified the asymmetry in convection surrounding the center of Invest 91C in the IR satellite imagery (Fig. __), with the region of the highest maximum reflectivity located just to the north of the TC center (Figs. __). All three simulations also highlight, with varying degrees of accuracy, the remnants of Invest 91C’s bent-back warm front that extends from the northwestern to northeastern periphery of the cyclone. The final position of the center of circulation is remarkably similar in all MP schemes (~42°N, 146°W) and is consistent with observations. The Kessler MP scheme produces the deepest surface cyclone, with a central pressure bellow 984 hPa (Fig. __). In contrast, the WSM6 and WSM3 MP schemes produce ~988 hPa cyclones (Figs. __). Unfortunately, there is insufficient observational evidence in this portion of the eastern North Pacific to prove that Kessler has overestimated the central pressure of Invest 91C compared to WSM6 and WSM3. All three MP schemes indicate that Invest 91C was a warm core cyclone at 700 hPa at 1200 UTC 1 November. This warm core extends above 500 hPa in each model simulation (not shown). The asymmetry illustrated in the 700-hPa wind field matches that of the convection, with the fastest winds consistently to the north of the cyclone center. The Kessler scheme continues to produce the deepest cyclone with the warmest core (~2°C). WSM3 produces the strongest 700-hPa radial temperature gradient and the fastest 700-hPa winds near the cyclone center. 9 Figure ____ depicts model derived outgoing longwave radiation (OLR) at 1200 UTC 1 November. WSM6 and WSM3 correctly identify the asymmetry observed in the cold cloud tops in the corresponding IR image (Fig. __). While the overall spiral structure of the OLR field is relatively similar between the two schemes, WSM3 has consistently warmer cloud tops. This is likely a result of how hydrometers are separated within each MP scheme. Only identifying three classes of hydrometers (vapor, cloud water/ice, and rain water/snow) could result in a reduction in condensate in the upper troposphere in WSM3 and an overall warmer solution. WSM6 yields the solution that most closely resembles the few available observations of the event. The WRF simulation that utilizes the Kessler MP scheme produces a highly unrealistic solution in the OLR field (Fig. ___). A vast expanse of cold cloud tops (> 90 W m−2 in some regions) covers the majority of the domain. These cloud tops do not correspond to regions of convection displayed in the simulated maximum reflectivity field (Fig. __). The reason behind the highly unrealistic solution observed in the in the OLR field lies in the assumptions embedded within the MP scheme itself. Only the autoconversion process can produce new precipitation particles in the Kessler parameterization. The autoconversion process will not occur unless a critical concentration of cloud droplets is exceeded. Fovell et al. (2009) suggest that the typical updraft speeds observed in TCs produce less condensation than continental convection, a smaller droplet concentration, and, therefore, fewer new precipitation particles when using the Kessler MP scheme. The unrealistic feature observed in the OLR field is likely an extensive anvil cloud. Figure ___, adapted from Fovell et al. (2009), displays vertical cross sections of the condensate distribution in TCs utilizing the Kessler and WSM3 MP schemes. Kessler exhibits considerably more condensate above 10 km than WSM3, likely manifesting itself as the spurious OLR field observed in the present study. 10 5. Summary and discussion The TT of Invest 91C occurred at ~40°N in the eastern North Pacific between 0000 UTC 29 October 2006 and 0000 UTC 2 November 2006. Despite the unusual location of TT, the physical processes associated with the cyclone’s transformation from an asymmetric, cold-core, EC into an axisymmetric, warm-core, TC are consistent with those found in previous studies. The present study supports the findings of Davis and Bosart (2003, 2004) and Hulme and Martin (2009a,b) that the precursor disturbance to TT is an EC that develops as an upper-tropospheric trough approaches a lower-tropospheric baroclinic zone (Figs. __). The EC progresses through the life cycle of a marine extratropical frontal cyclone described by Shapiro and Keyser (1990), developing a bent-back warm front on its west/northwestern side (Figs. __). As previously documented by Hulme and Martin (2009b), the bent-back warm front plays an important role in the TT of the EC. Convection along the bent-back warm front is believed to generate lower-tropospheric vorticity on the western half of the EC, intensifying cold air advection on the northern and western sides of the cyclone and helping to isolate the cyclone’s developing warm core. The diabatic redistribution of PV in the vertical along the bent-back warm front, upshear from the center of the cyclone, also helps to reduce vertical wind shear values over the center of circulation (Bosart and Davis 2004; Hulme and Martin 2009b). The bent-back warm front eventually separates from the center of circulation as the EC transitions into an axisymmetric, warm-core, TC (Fig. __). Invest 91C would ultimately make landfall as an unnamed TC along the northwest coast of Washington at ~1800 UTC 3 November 2006. The results of this analysis indicate that the structure of a cyclone undergoing TT is somewhat sensitive to the complexity of the MP scheme used in numerical simulations. Many similarities in the structure of Invest 91C are observed across the three simulations. All three MP schemes considered (WSM6, WSM3, and Kessler) 11 produce warm core cyclones by 1200 UTC 1 November with warm-core signatures evident above 500 hPa. All simulations also accurately capture the asymmetry in convection surrounding the cyclone center, with the deepest convection and the strongest lower-tropospheric winds predominantly on the northern side of the vortex. Subtle differences in the magnitude of the analyzed fields highlight the effects of the different MP schemes. The Kessler parameterization produces the deepest cyclone (~4 hPa deeper than either WSM6 or WSM3) with the warmest core at all analyzed levels. Despite having a cooler warm core, the steep radial temperature gradient observed in WSM3 produces the strongest low-tropospheric winds of any numerical simulation. The greatest disparity that results from the use of different MP schemes can be seen in the OLR field. The WSM6 parameterization produces the most realistic solution that is also the most consistent with the corresponding IR imagery. WSM3 captures the overall structure of the WSM6 OLR field, but fails to produce sufficiently cold cloud tops in regions removed from north of the cyclone center. The Kessler parameterization scheme produces the most unlikely solution, with the vast majority of the domain covered in an expansive cloud field. The simple representation of precipitation formation in Kessler causes condensates to remain in the upper troposphere once lofted. The result is the formation of the extensive and unrealistic anvil seen in Fig. __. The methodologies used in this study could be expanded upon to offer further insight into the structural differences that arise from changing the complexity of the MP scheme. Performing a similar analysis at higher resolution would provide a more detailed look at the structural disparities discussed about. Newer, more state of the art, MP schemes should also be incorporated into the investigation, specifically the Thompson MP scheme (Thompson et al. 2004). The same methodologies utilized here could also be applied to an entirely different EC developing over the continental United States. 12 Examining an EC over the continental United States would increase the likelihood of obtaining the surface observations, ground-based radar products, and upper-air data necessary to compare against WRF model output. 13