Data & Analytics

A Numerical Case Study of Secondary Marine Cyclogenesis Sensitivity to Initial Error and Varying Physical Processes

Description
A Numerical Case Study of Secondary Marine Cyclogenesis Sensitivity to Initial Error and Varying Physical Processes
Published
of 20
All materials on our website are shared by users. If you have any questions about copyright issues, please report us to resolve them. We are always happy to assist you.
Related Documents
Share
Transcript
  M AY  1999  641 C A R R E R A E T A L .   1999 American Meteorological Society A Numerical Case Study of Secondary Marine Cyclogenesis Sensitivity toInitial Error and Varying Physical Processes M ARCO  L. C ARRERA , J OHN  R. G YAKUM ,  AND  D A -L IN  Z HANG *  Department of Atmospheric and Oceanic Sciences, McGill University, Montreal, Quebec, Canada (Manuscript received 3 November 1997, in final form 1 June 1998)ABSTRACTSecondary cyclogenesis has been identified as a difficult forecast challenge. In this paper, the authors examinethe dominant physical processes associated with the predictability of a case of explosive secondary marinecyclogenesis and provide a better understanding of the large variability in the recent model-intercomparisonsimulations of the case. A series of sensitivity experiments, involving changes to the model initial conditionsand physical parameterizations, is performed using the Canadian Mesoscale Compressible Community Modelwith a grid size of 50 km.It is found that errors in the model initial conditions tend to decay with time, and more rapidly so in ‘‘dry’’simulations. The model fails to produce the secondary cyclogenesis in the absence of latent heating. Water vaporbudget calculations from the control experiment show that the surface moisture flux from 6 to 12 h is the largestcontributor of water vapor to the budget area in the vicinity of the cyclone center, and remains an importantmoisture supply throughout the integration period. During the first 12 h, these fluxes are crucial in inducinggrid-scale diabatic heating and destabilizing the lower troposphere, thereby facilitating the subsequent rapiddeepening of the storm. A secondary maximum in surface latent heat flux to the north and east of the primarymaximum acts to force the cyclogenesis event to the south and east of a coastal circulation center. When thesurface evaporation is not allowed, much less precipitation is produced and the secondary cyclone fails todevelop. Calculations of the potential temperature on the dynamic tropopause (i.e., 2-PVU surface) in the absenceof surface evaporation indicate a significantly damped thermal wave when compared with the control integration.This result for a case of secondary cyclogenesis differs from those generally found for large-scale extratropicalcyclogenesis where upper-level baroclinic forcings tend to dominate, and motivates the need for better physicalparameterizations, including the condensation and boundary layer processes, in operational models. The authorsspeculate that the different treatment of condensation and boundary layer processes may have been partlyresponsible for the enhanced variability in the simulation of this case in a recently completed internationalmesoscale model intercomparison experiment. 1. Introduction Along the east coast of North America, a special typeof cyclogenesis occurs frequently during the cold season(Miller 1946). Typically, a large-scale (mature) lowpressure system (i.e., a primary cyclone) propagatesslowly northeastward from the Great Lakes region. Coldair surges over the warmer waters of the Gulf Streamto the southeast of the primary low, acting to destabilizethe lower troposphere, while strong warm advection tothe north and east produces a significant low-level bar-oclinic zone along the coast. In this region, a secondary * Current affiliation: Department of Meteorology, University of Maryland at College Park, College Park, Maryland. Corresponding author address:  Mr. Marco L. Carrera, Departmentof Atmospheric and Oceanic Sciences, McGill University, 805 Sher-brooke St. West, Montreal, PQ H3A 2K6, Canada.E-mail: marco@zephyr.meteo.mcgill.ca cyclone, with a diameter of 500–1500 km, often formsin the wake of the preexisting primary cyclonic circu-lation. This type of cyclogenesis has been identified asa difficult forecast challenge (Kuo et al. 1995; Snyder1996). This has motivated the meteorological commu-nity to conduct the Fronts and Atlantic Storm-Track Experiment (Snyder 1996; Joly et al. 1997). Linked withthis forecast challenge was a need to clarify the role of upper- and lower-tropospheric processes in triggeringthe secondary cyclogenesis.Recently, an international mesoscale model intercom-parison effort known as COMPARE (Comparison of Mesoscale Prediction and Research Experiments;Gyak-um et al. 1995) was undertaken to understand furtherthe predictive capability of numerical weather predic-tion models on marine cyclones. One of its long-termobjectives is to choose cases of meso-  -scale phenom-ena from high-resolution field experiments and to per-form model intercomparison experiments in an effort toidentify systematic modeling errors (Chouinard et al.1994). The first case chosen for COMPARE was an  642  V OLUME  127M O N T H L Y W E A T H E R R E V I E WF IG . 1. (a) Traces of central SLP for the IOP-14 storm from COM-PARE models (dashed). (b) Six-hourly cyclone positionsstartingfrom0000 UTC 7 March 1986. For both (a) and (b) the RPN-preparedanalyses is given in thick solid lines. Arrow in (b) points to cyclonepositions at 24 h. Latitude–longitude lines in (b) are shown each 10  .F IG . 2. RPN-prepared analyses SLP (solid) at intervals of 1 hPaand surface temperature (dashed) at intervals of 5  C for 0600 UTC7 March 1986. The location of two buoys, 44005 and 44008, aregiven by    and   , respectively. Solid circles denote 45  N, 70  W;40  N, 70  W; and 40  N, 75  W. explosive marine cyclone off the North American eastcoast that occurred between 6 and 9 March 1986 duringthe concurrent field programs of the Canadian AtlanticStorms Program (CASP; Stewart et al. 1987) and theGenesis of Atlantic Lows Experiment (GALE; Dirks etal. 1988). Because of its development during the 14thintense observing period (IOP) of CASP, we term thisstorm as the IOP-14 storm. Twelve state-of-the-art lim-ited-area models from seven countries performed 36-hsimulations. For a detailed description of the variousmodels, see Gyakum et al. (1996).Figure 1 compares the simulated central sea levelpressure (SLP) traces and 6-hourly positions by theCOMPARE models as verified against the observedevent. The range of 12-h simulated central SLP valuesis about 7–8 hPa, which increases to 15 hPa at the endof the 36-h simulation (Fig. 1a). The 6-h positions of the low center exhibit a bifurcation of the tracks after18 h, with several models tracking the system too farwestward and inland; they recover thereafter to a moreaccurate position at 36 h (Fig. 1b). The largest vari-ability in cyclone positions occurs near 24 h, as indi-cated by the arrow in Fig. 1b.The bifurcation in cyclone tracks can be attributedpartly to the lack of high-resolution observations to re-solve three analyzed surface mesocyclones, as shownin Fig. 2 by the Recherche en Pre´vision Nume´rique(RPN) prepared SLP analysis. A time series of SLP andsurface winds from two buoys, 44005 (42.7  N, 68.3  W)and 44008 (40.5  N, 69.5  W) (see Fig. 2), supports theexistence of three mesocyclones in the analysis at 0600UTC 7 March 1986. In fact, none of the COMPAREmodels was able to reproduce the three circulation cen-ters (as shown in Fig. 2) at either 50- or 25-km hori-zontal resolution. All the COMPARE models repro-duced the continental low; however, most of the modelsproduced only the southeastern center offshore, missingthe coastal circulation. A few models missed the off-shore center altogether, producing only a circulationcenter along the coast. An examination of the detailedsurface SLP analysis reveals that the IOP-14 storm wasassociated with the southeastern circulation center with-in the surface trough. Some of the COMPARE modelsproducing the significant westward bias were found todeepen preferentially the middle circulationcenteralongthe New England coast, while those models possessinga more accurate track deepened the southeasternmostcenter within the trough after 18 h.  M AY  1999  643 C A R R E R A E T A L .F IG . 3. Time series of rms errors (dashed), as calculated againstthe RPN-prepared analyses, for COMPARE models: (a) Wind speed(m s  1 ) at 300 hPa, (b) temperature (  C) at 850 hPa, and (c) geo-potential height (m) at 300 hPa. Composite scores are given in thick solid lines. Figure 3 displays the time series of domain averagedroot-mean-square (rms) errors, as calculated against theRPN-prepared analyses, over the verification domaingiven in Fig. 4, for wind speed, temperature, and geo-potential height for each of the participants in COM-PARE. Specific features of concern are the increasingmagnitude and range of rms errors with time. Theresultsfrom Figs. 1 and 3 point to significant variability in theperformance among individual models in simulating acase of explosive oceanic secondary cyclogenesis.The objective of this paper is to perform a series of sensitivity tests, using one of the COMPARE models,in an effort to understand better the crucial physicalprocesses associated with the predictability of the sec-ondary cyclogenesis. This understanding may provideinsight into why the simulations from COMPARE areso variable. It should be mentioned that this storm hasbeen simulated by Mailhot and Chouinard (1989) usingthe Canadian regional finite-element (RFE) model witha grid size of 100 km, verifying against a coarse (300km) horizontal resolution objective analysis.Theirstudyfocuses, in part, on the role of the low-level jet and itscoupling with the moisture field during the rapid deep-ening phase of the IOP-14 storm. Here we use higher-resolution analyses and model simulations to aid in theidentification of important mesoscale structures (in thesea level pressure and surface latent heat flux fields),which are important to the predictability of the IOP-14storm. In addition, we examine the hypothesis of Kuoet al. (1991) that surface energy fluxes during the pre-conditioning period are important by means of a quan-titative quasi-Lagrangian moisture budget. The nextsec-tion describes the mesoscale model used for this study.Section 3 provides a brief synoptic overview of thestorm and outlines the experiment design. Section 4discusses the results of the sensitivity tests. A summaryand conclusions are given in the final section. 2. Model description and initialization a. MC2 model The model used for this study is the Mesoscale Com-pressible Community Model (MC2). The srcinal ver-sion of the MC2 model has its roots in the semi-La-grangian, semi-implicit, hydrostatic, primitive equationforecast model developed by Robert et al. (1985). Tan-guay et al. (1990) relaxed the incompressibility as-sumption inherent in the primitive equations, general-izing the semi-implicit algorithm to integrate the fullycompressible, nonhydrostatic Euler equations. A sum-mary of the MC2 modeling system is provided in ap-pendix A.Three different resolvable-scale precipitation param-eterizations to account for condensation under convec-tively stable conditions are used in the sensitivity ex-periments. The first one is a simple large-scale conden-sation scheme that removes moisture when the relativehumidity in a layer exceeds a specified value. Amajorityof the COMPARE models utilized such a scheme. Thesecond one is a predictive cloud-water scheme (Sund-qvist et al. 1989) permitting a more advanced treatmentof the mesoscale nature of clouds and precipitation sys-tems. A fractional cloud cover within the model grid is  644  V OLUME  127M O N T H L Y W E A T H E R R E V I E WF IG . 4. MC2 model integration domain with topography. Orography shown at contours of 500m. The verification domain is shown in thick solid lines. defined, allowing for the partitioning between precipi-tating water and cloud water. The third scheme followsManabe et al. (1965), removing excessive moistureabove a given relative humidity threshold. Unlike thefirst scheme, the effects of evaporation and freezing/ melting of precipitation are not included;theatmosphereis assumed to have zero carrying capacity and hencethe excess moisture immediately condenses and precip-itates. In each of the above schemes, the latent heatreleased is incorporated into the thermodynamic energyequation.The impact of three convective parameterizations(Kuo, Manabe, and Fritsch–Chappell) is examined.These schemes were chosen based on their use amonga wide spectrum of model performance in COMPARE.The Kuo scheme (Kuo 1974; Anthes 1977) defines aconvectively active layer as being a conditionally un-stable layer with a net positive moisture convergence(due to large-scale moisture convergence and surfaceevaporation). A fraction of this moisture is used to hu-midify the layer, while the remaining moisture con-denses and precipitates, heating the atmosphere in theprocess.In the version of the moist convective adjustmentscheme (Manabe et al. 1965), conditional instability isremoved by mixing adjacentlevels(coolingbelow,heat-ing above) to produce a lapse rate less than the dryadiabatic, the moist adiabatic, or a transitional combi-nation thereof. The scheme is triggered when condi-tional instability occurs with relativehumidity in agivenlayer exceeding a critical value in the presence of up-ward motion. The relative humidity in the layer is con-strained, after convective adjustment, to equal the valueat the bottom, and hence an upward transfer of moistureis required. Any excessive moisture condenses and pre-cipitates.The Fritsch–Chappell (1980) scheme assumes thatavailable buoyant energy, vertical wind shear, and ver-tical motion govern the evolution of deep convection.This scheme is well suited for grid sizes near 20 km,because the scheme assumes that only one type of con-vective cloud controls the vertical fluxes of heat andmoisture. b. Initialization All integrations are initialized at 1200 UTC 6 March1986 and integrated for 36 h. The initial conditions andsubsequent 6-hourly analyses were generated by the re-gional data assimilation system at the Canadian Mete-orological Centre (Chouinard et al. 1994). Additionaldata sources, including special buoy, ship, radiosonde,and aircraft dropsonde reports were obtained from theCASP and GALE observing networks and incorporatedinto the assimilation system.The assimilation cycle produced 6-hourly analyseson  M AY  1999  645 C A R R E R A E T A L .F IG . 5. RPN-prepared analyses SLP (solid) at intervals of 4 hPaand surface temperature (dashed) at intervals of 5  C for (a) 1200UTC 6 Mar, (b) 0000 UTC 7 Mar, and (c) 1200 UTC 7 Mar 1986.Note L1 denotes the primary continental low center, and L2 denotesthe IOP-14 storm. Latitude–longitude lines are shown each 10  . a polar stereographic projection (true at 60  N) with a50-km resolution in the horizontal shown in Fig. 4. Thisgrid comprises the model integration domain. The me-teorological fields of temperature, specific humidity,winds, and geopotential heights (hydrostatically bal-anced) are available on 44 isobaric levels, ranging from1050 to 100 hPa with a 25-hPa resolution, and includingthe 70-, 50-, 30-, and 20-, and 10-hPa levels. To avoidproblems associated with the lateral boundary condi-tions, a smaller domain (Fig. 4) over the CASP–GALEnetwork is used for model verification (Chouinard et al.1994). 3. Case description and experimental design a. Synoptic overview Figure 5 shows the RPN-prepared SLP analysis fieldat 12-h intervals for the period from 1200 UTC 6 Marchto 1200 UTC 7 March. At 1200 UTC 6 March, that is,the model initial time, a low pressure center (999 hPa,denoted by L1) was situated over southwestern LakeErie (Fig. 5a), having traveled eastward from the Ca-nadian Rockies in the previous two days. A large-scaletrough extended southward from the low center into theGulf of Mexico. At 500 hPa, the flow was characterizedby the presence of two troughs, one situated along theManitoba–Ontario border, the other centered in the OhioValley (Fig. 6a).By 0000 UTC 7 March, the low center had intensifiedand propagated eastward to a position over eastern LakeOntario (Fig. 5b). The IOP-14 storm formed in an areasouth of Long Island within the preexisting cycloniccirculation associated with the surface trough extendingsoutheast from the low center, as denoted by L2. Forthis reason, we consider this event asacaseofsecondarycyclogenesis. At this same time, at 500 hPa, the north-ernmost trough propagated eastward at a greater rateand the two short waves began to interact (Fig. 6b). Anarrow zone of high potential vorticity (PV) (Hoskinset al. 1985) values extended southeastward between thetwo troughs.Three circulation centers were present at 0600 UTC7 March (Fig. 2), with the southeastern center repre-senting the IOP-14 storm. The cyclone underwent itsmaximum deepening of 19 hPa/12 h to a position onthe southwestern tip of Nova Scotia at 1200 UTC 7March (Fig. 5c). Only a large-scale trough remained inthe position of the former continental system. Numerousobserving stations along the east coast of Canada andthe United States reported significant amounts of pre-cipitation. Also a secondary maximum in precipitationnear Cape Cod was associated with the coastal circu-lation center shown in Fig. 2. Reports of thunder andlightning from some maritime stations indicated con-vective activity (Mailhot and Chouinard 1989). The twoseparate troughs at 500 hPa had merged (Fig. 6c), andthe IOP-14 storm was located beneath the right entrance
Search
Similar documents
View more...
Related Search
We Need Your Support
Thank you for visiting our website and your interest in our free products and services. We are nonprofit website to share and download documents. To the running of this website, we need your help to support us.

Thanks to everyone for your continued support.

No, Thanks