USING THE THERMAL WIND RELATIONSHIP TO IMPROVE OFFSHORE AND COASTAL FORECASTS OF EXTRATROPICAL CYCLONE SURFACE WINDS

James B. Truitt

NOAA/National Weather Service Weather Forecast Office Juneau, Alaska Abstract

problem: they typically develop in data-sparse regions before they approach a coastal forecast area, making Land-falling numerical oceanic model cyclones predictions in the more midlatitudes uncertain confront and the forecastingthe forecaster of surface with a specificwinds more

article describes an alert system that helps the forecaster detect stronger synoptic-scale thermal difficult. Errors in surface wind forecasts can deeply affect coastal mariners and communities. This

gradients than otherwise expected and therefore identifies significant numerical forecast errors of cyclogenesis and associated surface wind.

A thermal-wind observation that is significantly stronger than the numerical model forecast value can indicate significant errors in the modeled cyclone evolution, typically including surface winds. The thermal wind observation is feasible in two steps: (1) an identification of regions suitable for the quasi-geostrophic assumption, and (2) a thermal-wind calculation that can be especiallycompared forto the the numerical analysis of model the thermal prediction gradients of thermal associated wind. with Our midlatitudeability to discern oceanic when cyclogenesis, and where to apply the quasi-geostrophic assumption has significantly improved over the last 20 years,

and this progress is briefly reviewed. Synoptic-scale regions of quasi-geostrophic flow can typically cyclonesbe identified moving using toward satellite the imagery, West Coast although typically other have data a synoptic-scale are useful to confirm region wherethe absence the geostrophic of approximationsources of strong is valid;ageostrophy and it arrives such as onshore deep convection before the and strong jet streaks. ageostrophies North Pacificassociated extratropical with upper

thermal wind to be compared to the numerical model value before the cyclone can generate stronger level jet streaks. This sequence of events during landfall permits wind shear-based estimates of the

than forecast surface winds in coastal regions. The thermal wind observation period has ended when the jet streaks are within subsynoptic range of the coastal wind profile site. Single significant- figure observations are sufficient to alert forecasters of possible errors in the numerical surface wind forecasts. An example is given using data from a rawinsonde and a Weather Surveillance Radar-1988 Doppler (WSR-88D) during a forecast shift at the National Weather Service Weather Forecast Office (WFO) in Juneau, Alaska. Based on the success of these results, suggestions for further research are provided.

Corresponding author address: James B. Truitt 4936 Hummingbird Lane Juneau, Alaska 99801-9230 E-mail: [email protected] Truitt

1. Introduction and occasional surface winds derived from satellite

scatterometer measurements. The upper-level data The forecasting of mid-latitude oceanic cyclones is winds,include Geostationary and observations Operational from the Environmental Aeronautical Satellite Radio, shortlyimportant before to operational moving into meteorologists a coastal forecast in coastal domain, regions. and (GOES) visible and infrared imagery, cloud-track GOES Such cyclones can develop in a data-sparse location Incorporated,Low observation Communication, densities Addressing are especially and Reporting evident themajor largest numerical departures model from busts the still model occur forecasts (McMurdie as early and betweenSystem (ACARS)(Benjamin the sea level and et al. the 1999). cloud tops of oceanic Mass 2004). In these cases, the forecaster needs to detect as possible. cyclones. A variety of satellite imagery is available, but Figure 1 indicates the locations of several forecast blockexcept the for detectiondata from of Microwave lower-tropospheric Sounding features,Units (MSU) and offices that routinely need to predict midlatitude oceanic furthermore,(Kidder and thisVonder concealing Haar 1995), cloud a shield cirrus can shield track can in cyclone evolution in the North Pacific Ocean. The cases detailed by McMurdie and Mass (2004) occurred near the southern boundary of the region shown in Fig. 1. Figure phase with the wave for a long time. Rawinsondes and 1 also indicates the locations of all wind profile sites thatdropsondes this data (Douglas density problem1990) are is typicallytypically not unavailable, solved by discussedalthough below. relatively more abundant at the surface except during specially funded research projects. Note Observations near oceanic cyclones are sparse, data include widely scattered ship and buoy reports, ACARS, since these observations are irregular (Businger et and the jet level than the intermediate layer. Surface al. 2001) and oceanic overflights typically use jet cruising levels for long distances. meteorologist can monitor coastal The verticaloperational wind

profiles in near real time from rawinsondes or WSR- 88D radars, but it may be difficultbetween at times the to observed identify significant differences

wind profile speed and and the model speed shearvalues. can Discrepancies be readily seen, in

in directional shear can be but important differences

difficult to detect visually data(Koch density 2001). can negatively impact The thick the layer computer of low models’ representation of a cyclone, and consequently limit the coastal forecaster’s

Situationalthe thick layerAwareness contains (SA). The problem is that that are crucial to the correctunknown values forecasting for fields of Fig. 1. cyclone evolution such as Location of Yakutat rawinsonde site (circle), Doppler radar sites (dark triangles) shown for the outer coast and offshore islands, a temporary wind profile site (open square), and Forecast154 National Offices Weather (open triangles). Digest Using the Thermal Wind Relationship to Improve Offshore and Coastal Forecasts

have to wait for the cyclone to be onshore before learning the baroclinicity. Forecasters in coastal regions sometimes part of a jet-front system reaching a WSR-88D site before the arrival of significant ageostrophic features closer to a how well the operational models performed. Large errors jet stream. are sometimes observed. landfalling Expanding cyclones on this is presentedthermal-wind here, approach where baroclinicity using the The forecaster can learn much about a cyclogenetic isVWP estimated and rawinsonde over thick layersdata, froman analysis the troposphere technique above for system if a thermal wind observation (TWO) becomes emphasizedavailable. inFor this example, article, the the thermal baroclinicity wind is a measurecan be conveniently estimated (Pettersen 1956). Also, as the boundary layer. These measurements are useful even values can be estimated by calculations using a vertical if only accurate to within one significant figure because of the horizontal temperature gradient. Thermal wind thethat thermal is sufficient wind toequation alert the and forecaster the conversion of possible of potential large errors in the numerical forecasts. Section 2 briefly reviews wind profile or a plan-view array of thickness values, provided the geostrophic component is significantly energyduring to midlatitude kinetic energy oceanic during cyclogenesis cyclogenesis. based Section on the 3 larger than the ageostrophic component. This condition explains how a quasi-geostrophic region can be identified isthe often data-sparse called the layer geostrophic associated approximation. with midlatitude oceanic This article describes a method to retrieve a TWO from Neimanlocations and near Shapiro these cyclonescategories, where along the with thermal analyses wind of model value for the corresponding location, then it is scale. In Section 4, we apply the conceptual model to associatedcyclogenesis. with If this stronger TWO subsequentis stronger than cyclogenesis a numerical and method is especially useful. A case study is presented in do not require measurement of surface pressure, and Section 5 to illustrate the usefulness of the technique, thereforesurface winds the than method the model can be forecast. applied These to the associations analysis of followed by a frequency of use evaluation in Section 6. Section 7 is a discussion with suggestions for further 2.research Thermal and SectionWind and8 gives Evaluation the conclusions. of oceanic cyclones in data-sparse regions. Cyclogenesis Our ability to discern when and where to apply the describedquasi-geostrophic how the approximationapproximation seemed has significantly valuable for calculatingimproved over thermal the windlast 60 values, years. but Forsythe also mentioned (1945) During the approach of oceanic cyclones, coastal thermalwind profiles wind isand a validsatellite indicator MSUs ofoffer cyclone the possibilityintensity, the of that upper air analysts experienced unexplained failures. retrievingretrieved data model-independent can be compared thermal to numerical wind modelvalues. values If the The quasi-geostrophic approximation has subsequently of thermal wind and therefore indicate the accuracy of the beenconvection, deemed strongly-curved “generally useful” geopotential (Neiman height and contours, Shapiro 1989)pronounced with exceptionsorographic thatbarriers, fit into the five boundary categories: layer, deep and modelof a lower-level forecast of geostrophiccyclogenesis, wind including from the an wind upper-level field. The thermal wind is defined as the vector subtraction equationjet streak to entrance retrieve and horizontal exit regions. thermal Neiman gradients and inShapiro cases (1989) and Koch (2001) have used the thermal wind geostrophic wind. g pu g pl involving upper tropospheric jet stream frontal zones. T is the geostrophic(1) theirNeiman associated and Shapiro temperature (1989) advectionsused single-station in the vicinity hourly of V V ( ) – V ( ) g wind, pu and pl are two pressure levels, and pu < pl can profiler observations to estimate thermal gradients and where VT is the thermal wind vector, V . VT shouldbaroclinic not zones. be assumed They elaborate. on the seminal article by be estimated from wind profile and sounding data if a Forsythe (1945), and identify cases where quasigeostrophy quasi-geostrophicwind may also be expressed approximation as: is made. Koch (2001) used mesoscale Using the definition of geostrophic wind, the thermal models and products from the WSR-88D, including a = — k X ∇ thermal-wind method applied to the Vertical Wind Profile f u l (VWP) data, to study a split front and the associated T 1 (2) convectiveforecasts and rainband evaluates in thethe usefulness southeastern of the United geostrophic States. V (Φ – Φf is) the Coriolis parameter, He compared the measurements to mesoscale model where Φ is the geopotential, assumption. His case study depicts a quasi-geostrophic and k is theVolume unit vector 32 Number along 2 ~the December z-axis 2008(vertical). 155 Truitt

u l Equation A schematic (2) can example be used of howfor calculating these equations thermal-wind might be thethat cyclone’sends with tracka cloud and pattern upper-tropospheric as depicted in Fig. features, 2 and values from model thickness fields, Φ - Φ . that these images agree with numerical model fields for appliedthe thermal to a windcyclogenetic is proportional system tois theshown thickness in Fig. gradient 2 (after including the size and shape of the cloud shield. Further Young and Grant 1995, Fig. 5.1.4a), where the magnitude of theassume numerical that GOESmodel cloud-trackdata to depict winds the deep and layer ACARS between also corroborate the numerical model solution. We only have and the length of the arrow would increase (decrease) that the data ingested by the model has been limited since as the thickness gradient increases (decreases). This the cirruscloud shieldshield developedand the ocean and surface.began to Thetravel problem in phase is figureof the alsovector, illustrates and that that its strengththe thermal is proportional wind is parallel to the to thickness contours, with lower values (cold air) to the left to simulate a variety of physical processes, including the with the perturbation. We are depending on the model magnitude of the thickness (thermal) gradient. While not a closed system, the kinetic energy of a cyclone Figure 2 also contains a callout that graphically applies derivesconditions from within the the available growing potential void of energyobservational released data. in howEq. (1), the andshear the of outlinedgeostrophic arrow winds is the can same be used vector to calculate as that shown within the thickness field. Therefore Fig. 2 depicts the numerical model provides an approximation of this how it might change, has diagnostic value including the rearrangement of the air masses (Reed 1990), and thatthe horizontal an increase temperature in the horizontal gradient. temperature This gradient, gradient and provides additional potential energy to a perturbed state hiddenprocess layeralong may with become the subsequent less representative wind field. of the real atmosphere, The model as timesimulation passes andof thethe void processes of observational in the that is available for conversion to kinetic energy (Carlson 1991). Assume This observation that we have of thickness a loop ofgradient satellite can imagery also be data persists through additional model runs. Therefore, compared to the numerical model value by the forecaster. observationalavailable from dataunderneath are still the sought cloud from shield, the they hidden can layer.show whether If thermal-gradient the potential energy observations available forthen cyclogenesis become

isconverted significantly to kinetic higher energythan the in model the cyclone, depiction. then If the a significantresulting surface amount winds of thatshould higher be consistent potential with energy having is

experience, such changes typically include higher wind higher kinetic energy than the model. Based on forecaster

speeds in the lower troposphere and at the surface. Cases also occur that are similar to the depictions in McMurdie and Mass (2004) where the subsequent cyclone evolution is too different from the numerical model solution to be usefullya representative compared. simulation of the potential energy The forecaster will prefer computer model(s) with Fig. 2. available for that conversion. If the values retrieved Schematic showing a leaf cloud (stippling) during thefrom forecaster observations will using subjectively Eq. (1) are strengthen significantly the higher model cyclogenesis (after Young and Grant 1995, Fig. 5.1.4a, p. 208) than the values predicted by the models using Eq. (2), with additional conceptual model details (after Holton 1992, the observational data are at a single location because the Fig. 6.5). Mid-level cloud E (hatching) has emerged from under solution.most important Such a cyclonecomparison mechanisms technique are is synoptic-scaleuseful even if cirrus shield F (stippling) between jet streaks J1 and J2 (arrow heads). Upper-level stream lines (long arrows) indicate flow betweenthrough short-wave the streamlines trough is shownaloft (dashed as heavy line). dashed Surface lines, fronts with processes (Parsons and Smith 2004). (conventional notation) show an inflection point. Thickness The literature has long mentioned cases where Eq. a thermal wind vector (outlined arrow) consistent with the (1) was applied to actual winds with a “lack of success” local thickness gradient at the vectors’ base. Figure callout (Forsythesources of 1945).error would Such befailures from resultthe presence from sources of one orof moreerror illustrates the relationship of the thermal wind (outlined described by Neiman and Shapiro (1989). The largest arrow)156 National to the geostrophic Weather Digest winds at 850 and 500 mb. Using the Thermal Wind Relationship to Improve Offshore and Coastal Forecasts

b. Scale of strong jet streak ageostrophies of the five categories of strong ageostrophy. Additional A useful measure of the validity of the geostrophic errors should result from the application of Eq. (1) in place of the complete thermal wind equation (Neiman and as the ratio of the characteristic scales of ageostrophic 0 Shapiro 1989), but consistent with Forsythe (1945), such approximationacceleration to isthe the Coriolis Rossby acceleration: number (R ), which is defined errorsthat checks should the be validity an order of of the magnitude geostrophic smaller. approximation Therefore the forecaster would benefit from a diagnostic method R U fL— for3. theRegions observed Containing wind profiles. Strong Ageostrophy 0 where U is the velocity scale, f is the Coriolis parameter,(3) Considering thermal wind observations in real-time and L oceanic cyclone that has the associated baroclinicity, yet is the horizontal length scale. The Rossby number wherebegins theby first actual seeking winds to have identify a geostrophic a sector componentwithin the is typically of order 0.1 for midlatitude synoptic-scale systems When (Holton R 1992) and of order 1.0 in the presence of canjet streaks be made, on subsynoptic although the scales thermal (Bluestein wind calculation1993). is section shows that the strong ageostrophies near such 0 significantly larger than the ageostrophic component. This ~0.1, the quasi-geostrophic approximation an adjacent sector suitable for the geostrophic assumption cyclones are sufficiently limited in spatial scale, such that limited to one significant figure accuracy. A comparison of the results of Eq. (1) using wind observations, with Eq. (2) a.and Synoptic-scale contributing regions to cyclogenesis containing can strong be identified. ageostrophy using Actual model measurement thickness fields, of Rcan is reveal not practical; large errors however, in the model estimate of available energy. were temporarily available 0

0 validity of the quasi-geostrophic assumption can be numerical model fields of R The Neiman and Shapiro (1989) exceptions to the Rat WFO Juneau, and were applied to the thermal wind method. Cases of R0=0.2 occurred, along with one case of 0 identified using data and techniques already available to =0.3. These cases have been used to estimate a threshold operational meteorologists: (1) Deep convection can be value for the significance of the ratio of observed thermal detectedalong with using other satellite sources imagery of error or such WSR-88D as data corruptionreflectivity wind magnitude over model thermal wind magnitude. If data. (2) Strongly curved geopotential height contours, the ratio is above 1.3, then the measured thermal wind value is significantly higher than the model value, with due to velocity aliasing, should be detectable using WSR- inapplicableincreasing values and having therefore increasing cannot significance. be used to identify 0 barriers88D velocity on the data outer displayed coast can with produce the VWP barrier root jets mean that In subsynoptic regions where R ~1, Eq. (1) is squared error (RMSE). (3) Pronounced orographic Furthermore, the vertical coupling of ageostrophic winds associatederrors in withthe isotachmodel maximaestimate can of resultavailable in the energy. depth terrainare not cansignificant typically above be neglected the terrain by not level using (Parish data 1982).below of strong ageostrophy extending below the jet-stream (4) Boundary layer effects over oceans and shallow coastal surface wind must still account for a transition through the 850-mb level; however, a method for forecasting level (Bluestein 1993). Therefore, Eq. (1) should not be theapplied cyclogenetic to wind profile region dataone fromseeks beneath to sample the for jet thermalstreaks. the boundary layer. gradientsThese ageostrophies and the geostrophic are contained approximation within a subset can beof The remaining exception to the validity of the quasi- applied to the remainder of the region, as described below andgeostrophic the existence approximation of jet streak (Neiman entrance and Shapiroand exit 1989) regions is from upper-level jet streaks. The location of a jet stream 4.with Conceptual a conceptual Modelmodel. and Application are easily identified using satellite imagery (Kidder and Vonderexists within Haar synoptic-scale1995). Note that distances the jet ofstream the cyclogenetic is intrinsic a. Cyclogenetic regions: quasigeostrophy vs. ageostrophy to cyclone evolution (Reed 1990) and therefore routinely topic of the strong ageostrophy associated with jet streaks An oceanic cyclogenetic system is depicted baroclinicity field one seeks to sample. Consequently, the conceptual model of cyclone development to identify a is emphasized in this article. We will now use an existing schematically in Fig. 2. The cirrus shield, depicted with the stippled area F, is commonly identified with infrared Volume 32 Number 2 ~ December 2008 157 sector where the geostrophic approximation is useful. and water vapor imagery. Individual jet streaks along the Truitt jet stream can be located using loops of such images, but forecaster blends experience with a conceptual model solutions,of cyclogenesis and observations that includes including the conversion thermal of wind. potential The thevertical precision circulations of streak associated identification with is better jet streaks transverse are of the flow than along it. The details of the transverse/ in the presence of these jet streaks on subsynoptic scales energy to kinetic energy (Bluestein 1993). Usually, the 0 beyond the scope of this article. R ~1 should be occurring correspondingthermal wind numericalestimate calculatedmodel solutions from calculatedwind profiles with assumed near where the cirrus shield area F is depicted with Eq. (1) agrees to one significant figure with the (Bluestein 1993). Similar strong ageostrophies are due to the complexity of the jet streaks that should soon Eq.thermal (2), which wind showsis associated that the with numerical stronger forecast cyclogenesis represents than as adjacent cloud shield E, but will be harder to identify the energynumerical conversion model solution, about to occur.and stronger However, cyclogenesis a stronger cannotcouple. correctlyThe horizontal draw extentR isopleths of the strongfor various ageostrophies levels on the forecaster must adapt his or her analysis, as illustrated is not known to a subsynoptic-scale resolution, i.e. we is associated with stronger surface wind. In such a case, 0

Fig.subsynoptic 2. However, scale the precision jet streaks by forecastershave locations using relative satellite to 5.in theForecast following Shift example. Case Study the cyclogenetic system that are routinely identified to whole cyclone, and this is consistent with model studies thatimagery. show Such the ageostrophicfeatures exist circulations on a smaller associated scale than with the illustrates how the forecasting of winds associated cyclones are typically an order of magnitude smaller with This an caseoceanic study cyclone from 11during January coastal 2002 approach at WFO Juneaucan be improved with a baroclinicity calculation that provides assume the modeled cyclones accurately represent actual than the geostrophic circulations (Anthes 1990). Let us circulations would include the jet streaks that the an alert. TWOs were calculated with the PAYA (Yakutat) cyclones. In Fig. 2 the smaller scale region of ageostrophic rawinsonde and the PACG WSR-88D wind profiles. The WSR-88D site was temporarily out of service when a aforecaster much larger identifies scale, in the the thermal western wind part equationsof the cyclone. can be measure of baroclinicity was first needed. The PAYA Because the geostrophic circulations take place on rawinsondeforecasters issued was first amendments used by the based forecasters on these for thedata, wind- and shear calculation and comparison with model data. The applied usefully. Assume that that the cyclone in Fig. 2 is in the North Pacific and is approaching a West Coast WSR- then began to receive WSR-88D VWP data, which they 88D or wind profiler. Such a location under the cirrus used for updated evaluations. The wind speed units used shield is useful because of: (1) the obscuration by the in this section-1 for the public-1 forecasts and verification are cirrus shield, (2) the role of the thickness field in cyclone mph (m s ), while kt (m s ) were employed for the marine evolution, and (3) the jet streaks are still a synoptic scale forecasts and verification.. Table 1 shows the chronology distance offshore from the wind profile sites. a.of Real-timeevents that vs. occurred hindsight during calculations the forecast shift. The flow becomes more ageostrophic as the jet streaks satellitemove within imagery subsynoptic-scale or cloud motion range winds of the to trackwind profilethe jet site, and the forecaster can anticipate this by using GOES time Two during thermal the forecast wind comparisons shift; secondly are aprovided more stringent in each steaks synoptically. Therefore, when the jet streaks are case. First is an initial crude comparison used in real- within subsynoptic range of the data site the TWO period has ended. A tabular history of the TWOs is useful. For comparison is applied later in hindsight. example, data from the WSR-88D site PACG (Biorka Island) During the forecast shift, the magnitude of the thermal (Fig. 1) is recorded along with the application of Eq. (1) by wind calculated with Eq. (1) from the 850-500 mb PAYA and PACG winds were compared with the magnitude of the aheada computer of these script west at coast WFO land-falling Juneau. The strong forecaster ageostrophies can then review a prior period of wind data collected sufficiently far measurementsthermal wind calculated were the lowest with Eq. available (2) from winds the due1000-500 to the mb layer Aviation (AVN) data. The 5000 ft (1524 m) PACG b.that Forecasting the quasi-geostrophic oceanic cyclogenesis approximation and surface is valid. wind boundary layer constraint. Hindsight calculations using A diagnosis of cyclogenesis can be made by blending the AVN 850-500 mb layer are provided. satellite data, conceptual models, numerical model

158 National Weather Digest Using the Thermal Wind Relationship to Improve Offshore and Coastal Forecasts b. Forecast shift briefing

A GOES IR animation ending with the image at 0000 UTC 11 January 2002 (Fig. 3) indicated cyclogenesis and a northward track. The 0000º UTCº National Center for amongEnvironmental the models Prediction were small, (NCEP) but surface the model map of (Fig. choice 4) showed a surface low at 53 N, 143 W. The differences by the prior shift was the 1800 UTC 10 January 2002 run of the AVN Global Model and had therefore guided the forecast package issued to users. Output from this model run at 6 h, 12 h, and 18 h is shown in Figs. 5-7. Figure 5 shows AVN thermal wind values available closest to PAYA and PACG, as well as the corresponding model thickness field at 6 h. Figure 6 shows the same model features, with the addition of the surface pressure field and the low that is tracking north at 12 h. Figure 7 shows the same model features as Fig. 6, except for the thermal wind values are omitted as they are no longer used for comparison, at 18 upper-tropospheric cyclogenesis with weak development Fig. 3. h. Forecasters believed that the satellite imagery showed in the lower troposphere, but the view was blocked by the Capital L denotes approximate location of sea-level pressure Infrared GOES image for 0000 UTC, 11 January 2002. extensive cloud shield (Fig. 3). Another problem was that minimum. the PACG WSR-88D site was out of service until 0308 UTC. The 0000 UTC PAYA rawinsonde was available but needed evaluation for usability with the geostrophic assumption. Therefore, the validity of the geostrophic assumption first neededTime (UTC)to be established. Event

1800 Last AVN model run on 10 January 2002 0000 Rawinsondes are completed, Fig. 4. map time Zone forecast issued, including Yakutat 0020 Marine forecast issued Forecasting crew change and notes started 0030 - 0100 ~0145 PAYA rawinsonde data are used for calculation of ther mal wind, Eq. (1) Wind Advisory issued for Yakutat zone 0206 Marine forecast is amended 0211 0308 PACG data begins to arrive, including VWP 0313 PACG VWP data are used for calculation of thermal wind, Fig. 4. amendmentsEq. (1) ~0330 PACG WSR-88D data diagnosed a validation of the 0000 UTC, 11 January 2002 theNCEP pressure surface trough, analysis. and Solid capital contours L’s denoteare isobars approximate (mb). Dashed locations line ofindicates 0600 AVN model verification time-1 (Fig. 6b) 0943 PAYA first 45 mph (20 m s ) gust sea-level pressure minima. Standard 1200 Ship WCZ6534 verifies Gale -1 Table 1. symbols denote frontal boundaries. 1242 PAYA gust of 46 mph (21 m s ) PAYA rawinsonde site is shown as open Volume 32 Number 2 ~ December 2008 159 Chronology of forecast shift 11 January 2002 circle. Truitt c Applicability of the geostrophic approximation

. that the flow over the coastal region and adjacent ocean was quasi-geostrophic. The satellite imagery at 0000 UTC Theº satelliteº imagery showed a baroclinic comma did not show convection along the outer coast to the NNE upstreamcloud with to an the upper-level south, but inflection not downstream point in thetoward vicinity the through E of the cyclone. Based on the conceptual model, of 54 N, 145 W (Fig. 3). Jet stream signatures extended the satellite imagery, and experience using the PAYA data model of jet streaks with ongoing cyclogenesis, we assumed (Truitt 1994), we decided that the baroclinicity could be coast including PAYA and PACG. Based on the conceptual calculated using the PAYA wind shears and Eq. (1). The PACG WSR-88D became available with the 0308 UTCageostrophies volume scan would (Fig. not 8). reach Based the on outer the AVN coast cyclone until manytrack (Figs. 6 and 7), the forecaster assumed that jet streak

werehours lesslater. than 4 kt, which is consistent with a near- The RMSE values for the winds in Fig. 8 (not shown)

homogeneous wind field at all levels. We therefore assumed the PACG data shown in Fig. 8 to be quasi- d.geostrophic. PAYA rawinsonde

thermal wind was compared to the corresponding model The solution 0000 UTC to establish PAYA rawinsonde whether cyclogenesis 850-500-mb should layer

generally follow the AVN prediction.-1 The PAYA thermal winds using Eq. (1) was 52 kt (27 m s ). The AVN thermal- wind-1 magnitude retrieved with Eq. (2) was 25 kt (13 m ) and this value is overlaid on Fig. 5. The magnitude of Fig. 5. AVN 6-h forecast of 1000-500 mb thickness contours the thermal wind at PAYA was therefore roughly double (dashed, dam) and thermal wind vectors (kt), valid at 0000 what the model predicted. UTC 11 January 2002. Locations of PAYA rawinsonde site As of 0000 UTC, the PACG WSR-88D was not yet (open circle) and PACG WSR-88D (dark triangle) are included. reporting. The forecaster used three factors to subjectively estimate that the thermal wind near PACG was about half again what the model guidance showed: (1) The PAYAAccordingly, thermal the wind numerical observation model had thermalusefulness wind for a values wide areaneeded including to be factoredPACG, but in diminishing more for increasing with distance. distances (2)

from the PAYA data site. (3) The satellite imagery was e.used PACG to WSR-88Destimate the VWP size of the cyclogenetic system.

The VWP from PACG (Fig. 8) first became available just after 0300 UTC, 11 January 2002. This was between the times of the AVN 6 and 12 hour forecasts valid at 0000 UTC and 0600 UTC 11 January 2002. Therefore, the magnitude of the thermal wind calculated with Eq. (1) Fig. 6. AVN 12-h forecast of sea-level pressure (solid, mb), was compared with the mean value of the AVN thermal 1000-500 mb thickness contours (dashed, dam), and 850-500 wind magnitudes. We used the VAD (Velocity Azimuth mb thermal wind vectors (kt), valid at 0600 UTC 11 January 2002. Capital L indicates location of model sea-level pressure Display) wind at 5,000 ft (1524 m) as an approximation minimum. Locations of PAYA rawinsonde site (open circle) and for the 850 mb wind, and the VAD wind at 18,000 ft (5486 PACG160 WSR-88DNational Weather(dark triangle) Digest are included. m) as an approximation for the 500-mb wind in Eq. (1). Using the Thermal Wind Relationship to Improve Offshore and Coastal Forecasts

January 2002. The forecast had no other wind increase overnight. The marine amendment transmitted at 0206-1 UTCWCZ6534 read “southeast winds increasing to 40 kt (21 m s ) Thursday evening.” At 1200 UTC 11 January 2002, ship-1 reported winds southeast at 40 kt (21 m s ) (Fig. 7). In both cases, the TWO had improved the surface g.wind Hindsight forecast. computations

made during a forecast shift: a comparison of thermal The case study above described the calculations

wind values retrieved from 850-500-mb observed winds with the thermal wind values calculated from the AVN Fig. 7. As in Fig. 6, except 18-h forecast valid at 1200 UTC 11 1000-500-mb thickness. In hindsight, a comparison of January 2002. Note the inclusion of the location of the ship thermal wind values retrieved-1 from 850-500-mb PAYA report from WCZ6534 (large black dot). winds (52 kt, or 27 m s ) with the thermal wind values calculated-1 from the AVN 850-500-mb thickness (19 kt, or Based on these winds observed at 0313 UTC,-1 the thermal 10 m s ) yields a ratio of 2.7. A comparison of thermal mwind s at PACG was found to be 39 kt (20 m s ), while the wind values retrieved-1 from 850-500-mb PACG winds (39 AVN-1 thermal-wind magnitude for PACG was 25 kt (13 kt, or 20 m s ) with corresponding-1 AVN 850-500-mb experience,). This the corresponds stronger cyclogenesis to a ratio wouldof 1.5 resultbetween in peak the thermal wind (19 kt, or 10 m s ) yields a ratio of 2.0. measured value and the model value. Based on forecaster Note that using the AVN 850 mb level instead of 1000 surface winds near PACG nearly 50% stronger than those f.predicted Amendments by the issued AVN. and verification received

s The Yakutat Public Zone from 0020 UTC 11 January 2002Advisory-1 stated was “winds issued becoming for the Yakutatsoutheast Zone to 25 and mph the (11 text m ) by the morning (of 11 January).” At 0211 UTC, a Wind was amended to “Southeast-1 winds increasing with gusts to 50 mph (22 m s )” for the overnight period. At 0943 UTC, the winds at PAYA-1 became southeast and increased to 24 mph (11 m s ). From 0953 UTC-1 to 1153 UTC four peak wind values of 45 mph (20-1 m s ) occurred, and the highest wind of 46 mph (21 m s ) occurred at 1242 UTC and 1253 UTC. marine observation was to become Marine forecasts received corresponding amendments at 0206 UTC. Only one available in the eastern Gulf of Alaska through 1200 UTC (Fig. 7) and the marine wind amendment for this location was based on a more conjectural use of the 0000 UTC PAYA TWO. The subsequent PACG TWO caused the forecasters to become more confident that the previously transmitted 0206 UTC marine amendment was appropriate for Fig. 8. VWP from PACG for 0308-0358 UTC 11 January 2002, the eastern Gulf of Alaska closer to PACG. The original as displayed on the Advanced Weather Interactive Processing marine forecast for this location, issued at 0020-1 UTC, read System (AWIPS) at the NWS Juneau WFO. Standard wind “southeast winds less than 20-1 kt (10 m s ) increasing barbs are shown in height scale for each 1000 ft (305 m). to southeast 25 kt (13 m s ) Thursday evening,” with Second column of wind barbs depicts VWP ending at 0313 UTC used for shear calculation. The veering of wind is evident for a Thursday evening defined as 0300 UTC to 0900 UTC 11 Volume 32 Number 2 ~ December 2008 161 large depth. Truitt

-1 mb reduced-1 the thermal wind from 25 kt (13 m s ) to 19 6) Noapproaching cyclone, atbased the time on of the application available of thehardcopies thermal kt (10 m s ), which increased the ratio to 2.7. The results of model analysis and GOES IR. A cold front was 6.validate Rates the of baroclinicity Occurrence alert issued during the shift. wind method. The ratio of thermal wind over model values was 1.33. No documentation was found that indicated whether the forecasts differed from Case studies were logged during an early period (cool model depictions, or whether verification became seasonpredictor of of 1999-2000) surface winds in forthe any techniques cases where development. landfalling available. The method should not have been applied During this period, the forecasts used the TWOs as a (one case). statistical analysis, but subjective comparisons with the features such as cold fronts might result in a significant The number of cases here is too small for formal ofwind the event. thermal These wind cases method are identifiedwas based and on includeda statistical for the purpose of completeness. This broad application subsequentFailures have seven subsequently years are occurredpractical. but The not first for three the categories have typical relative frequencies of occurrence. study (Truitt 1994). The case studies typically did not have occurred when winds, stronger than depicted by differentiate between the marine forecasts and the reasonsthe numerical given models,in 4, 5 occurredand 6. Theat high subsequent elevations failures in the forecasts for communities near sea level. The cases were classified into the following groups: interior mountains. 1) The TWO values were significantly higher than Failures such as (4) should sometimes still occur. The the numerical models (similar to the above case thermal wind method only samples the 850-500 mb layer, study). The surface winds were forecast and verified andthe numericaldoes not identify model upper values, tropospheric and they may features. result fromSkill as significantly stronger than numerical model islayers not claimed with stronger for cases thermal where wind the TWO values is eitherweaker above than predictions (two cases). the forecasters used the model guidance and standard 2) wasThe usednumerical to choose model the solutionsmodel with had stronger significantly winds or below the 850-500 mb sample layer. In (4) above, different wind forecasts. The thermal wind method techniques successfully. and did verify, including a High Wind Warning (one Cases where TWOs are stronger than the model values case). windscan have can actual occur incyclone locations evolutions that seem that inexplicable differ from using the model solutions in a variety of ways. The stronger surface 3) The thermal wind agreed with numerical model thevalues forecaster and the did documentation not expect cyclogenesis, then ended but (nine this the model solutions. Such cases are hard to classify, but cases). In two of the cases there is indication that 7.are Discussionroughly similar and to McMurdieProposed and Future Mass (2004).Research

detail is difficult to reconstruct. solutions and explosive cyclogenesis occurred as midlatitude oceanic cyclone was sampled for a thermal 4) The thermal wind was weaker than any model wind In calculation,this study thefrom data-sparse which the forecasterslayer associated had inferred with a applied the model solutions presumably using that the cyclogenesis was to become stronger than verified with ship data. The forecasters successfully were updated, and the updates monitored, with the satellite imagery and other standard tools. The expectationpredicted by that operational an increase computer in wind models. speeds The would forecasts reach method failed (one case). obvious convection after a log entry of ‘could these 5) The thermal wind method was applied during the surface. and forecast winds were raised higher than What This the simple operational method community should help needs provide most isdirection for the cells be severe.’ The models depicted cyclogenesis numericalto the research model and performance numerical tomodeling improve communities. enough for resulting wind advisory was claimed based on one numerical model solutions. The verification of the the research community are provided, including some for such methods to be obsolete. Therefore, suggestions to gust but the log has ambiguity. The method should 162 notNational have Weatherbeen applied Digest (one case). future theoretical work. Using the Thermal Wind Relationship to Improve Offshore and Coastal Forecasts

For example, the numerical modeling community could be above the boundary for the purposes of the thermal research the creation of model ensembles that emphasize a range of thermal wind values and their subsequent wind tool, yet weak orographic influences have probably beenfailures detected that have at PACG been even observed during in onshore the following flow. seven cyclone evolutions. The theory of extratropical cyclones The case studies from 1999-2000 did not show method (Hoskins 1990) extends far beyond the method presented winds stronger than depicted by the numerical models in this article, albeit the method is convenient for field years. These subsequent failures have occurred when operations. than the models occurred at high elevations as reported altered Based cyclone on the literature evolution (Carlson when there1991; isReed stronger 1990) were forecast, but the only winds significantly stronger we can typically expect stronger winds, or a significantly comparison tool is designed to calculate thermal wind byin aanemometers cyclone’s winds in the aloft interior that do mountains not extend near downward Juneau. fromthermal observations wind in the and cyclogenetic compare region. it with theThe numericalTWO and throughIn such cases the boundary the method layer seems until to reaching predict an topography increase experimentally that we can improve our forecasts with model solution. Tests at the WFO Juneau demonstrate above about 4000 ft (1219 m). future work should investigate the mechanisms behind Cases where the TWO is weaker than the numerical thisthe relationship,tool. Consequently, as well asthe the reasoning reasons for was the right observed and model values cannot be usefully applied. It is possible thermal wind values being stronger than the numerical that a strong baroclinic layer extends below 5000 ft (1524 toolm) that to failthe tonumerical detect cases model where depicts the as numericalbeing within model the 850-500 mb layer. If this is so, it is also possible for the assumptionsmodel solution. but introduces error, especially when the The TWO and comparison tool uses several simplifying surface winds will not be strong enough. The persistent void of observational data offshore observedwith the values assumption are stronger that the than cyclone the model evolution values: will(1) may require soundings for significant improvement of Thehave existing little change numerical except model for the results subjectively are typically determined applied numerical model performance. Dropsondes (Douglas increases in wind speeds in the lower troposphere 1990) could be placed offshore according to the conceptual model and synchronized for numerical model runs. The TWOs could also be calculated using winds at lower levels areand unknownthe surface. and (2) the The result increased is the assumption thickness gradient that no than is practical at sites like PACG, due to the shallower may have several unknown causes. However, they boundary layer over the ocean. Rapid changes are occurring in the field of Unpiloted Aircraft Systems (UAS) differentiationand most of the exists additional that may terms affect should cyclone be practical evolution. to that should lower the cost per sounding. Satellite MSUs (3) The complete thermal wind equation is not used, can also retrieve data through clouds (Kidder and Vonder Haar 1995). A comparison would be made between two implement (Neiman and Shapiro 1989). (4) The tool applications of Eq. (2): the first using model thickness does not account for boundary layer complexities. These fields, the second using a field of thickness values from u l include friction, stratification, baroclinicity, turbulence, the MSU soundings. This second application of Eq. (2) alongthe rapid with development their interactions. of oceanic In addition, extratropical the sensible systems would While use finite-differencing the conceptual model to obtain presented the field above of Φ has-Φ and latent heat fluxes in the boundary layer can “fuel” broadfor locations operational according applicability, to the conceptual the following model. limitations boundary(Uccellini 1990).layer, yet the associated method must account for The processes data used that for Eq. transport (1) need momentum to come from downward above the theshould data be site noted: if the cyclone (1) The is complexmaking landfall terrain to of the western south, North America often generates strong ageostrophies at throughthe surface the roughnessboundary layer changes to the from surface. water toIn landaddition, and or if the cyclone is following a coastal track from the south. the coastal wind profile sites are typically located where (2) The exact location of the jet streaks and associated operational meteorology indicates the boundary layer ageostrophyback with height can earlierbe difficult than toexpected determine. and are Cases believed have a deeper boundary layer. For example, experience in beento be observed associated at WFO with Juneau ageostrophic when the frontal wind transverse began to over the ocean is below 3000 ft (914 m). However, tests arrival of ageostrophies may be associated with mature using the PACG WSR-88D VWP show higher RMSE and circulations as described by Koch (2001). This early obvious exceptions to homogeneous flow through 4000 ft (1219 m). An altitude of 5000 ft (1524 m) is assumed to occlusions as definedVolume in32 theNumber Norwegian 2 ~ December conceptual 2008 model 163 Truitt

of cyclone evolution. (3) Measurements at a data site model The value,TWO is then compared at least to one the ofnumerical three model-relative model value. can correspond to a preceding baroclinic wave. (4) The If the observation is significantly stronger than the thermal wind for the 850-500-mb layer should also be thermalcalculated wind in two and or baroclinicity more layers, could such be as limited 850-700 to mb the outcomes will occur: (1) raised wind speeds in the and 700-500 mb, to help identify cases where the strong lower troposphere, (2) higher surface wind speeds, and (3) Forsubsequent operational cyclone meteorologists, evolution too thedifferent thermal from wind the 8.upper Conclusion layer, without extension to the boundary layer. model for operationally useful comparison.

methodpredict can numerical alert the model forecasters errors of should significant help numerical provide thermal wind as an elegant instruction tool, but much less model forecast errors. That such a simple approach can Historically, meteorologists have thought of the community, the thermal wind method described, direction to the research community. For the research as a useful analysis tool. The geostrophic approximation has a long history of failures (Forsythe 1945; Doswell emphasizes the need for sounding data (from satellites 1991), and strong ageostrophies are significant features of microwave sensors or aircraft) within oceanic midlatitude extratropical cyclones. However, the identification of such Authorcyclogenetic systems. scaleageostrophies region within has become the cyclogenetic more practical system (Neiman suitable and for Shapiro 1989) and allows the identification of a synoptic- Jim Truitt for assuming quasigeostrophy includes a low data-density volumethe quasi-geostrophic under the cloud approximation. shield that travels The region in phase suitable with is a Lead Forecaster at NOAA/National Weatherlong-term Service goals include Forecast improving Office in the Juneau. wind forecasting He earned of a B.S. in Meteorology from Texas A & M University. His the cyclogenetic wave. extratropical cyclones in the North Pacific. References

prediction of cyclone development with limited-area equipped aircraft for operational forecasting Anthes, R. A., 1990: AdvancesExtratropical in the Cyclones:understanding The Erikand Douglas, M. W., 1990:Bull. TheAmer. selection Meteor. andSoc. use of dropsonde- Palmén Memorial Volume, fine-mesh models. applications. , 71, 1746-1756. C. W. Newton and E. O. Bull. Amer. Holopainen, Eds., Amer. Meteor. Soc., 221-253. Forsythe,Meteor. G.Soc., E., 1945: A generalization of the thermal wind equation to arbitrary horizontal flow. Benjamin, S. G., B. E.Wea. Schwartz, Forecasting R. E. Cole, 1999: Accuracy of 26,An 371-375. Introduction to Dynamic Meteorology. ACARS wind and temperature observations determined by collocation. Observations, and 14, 1032-1038.Theory of Weather Holton, J. R., 1992: Systems. Vol 2, Synoptic-Dynamic Meteorology in the 3d. ed. Academic Press, 511 pp. Bluestein,Midlatitudes, H. B., 1993: Extratropical Cyclones: The Erik Palmén Memorial Hoshkins,Volume, B. J., 1990: Theory of extratropical cyclones. Oxford University Press, 594 pp. C. W. Newton and E. O. Holopainen, Eds., Amer. Businger, S., M. E. Adams, S. E. Koch, and M. L. Kaplan, 2001:Mon. Meteor. Soc., 64-80. Satellite Wea.Extraction Rev. of geopotential height and temperature Meteorology: An Introduction. structure from profiler and rawinsonde winds. Kidder, S. Q., and T. H. Vonder Haar, 1995: , 129, 1729-1739.Mid-Latitude Weather Systems. Academic Press, 466 pp. Carlson, T. N., 1991: Wea. HarperCollins Academic, 507 pp. Koch,Forecasting, S. E., 2001: Real-time detection of split fronts using the application of hodographs to forecasting severe mesoscale models and WSR-88D radar products. Doswell, C. A., III,Natl. 1991: Wea. A Digreview for forecasters on 16, 35-55. 164 National Weather Digest thunderstorms. ., 16(1), 2-16. Using the Thermal Wind Relationship to Improve Offshore and Coastal Forecasts

Wea. Forecasting, xtratropical Cyclones: The McMurdie, L., and C. Mass, 2004: Major numerical forecast Reed,Erik R. PalménJ., 1990: AdvancesMemorial inVolume, knowledge and understanding failures over the Northeast Pacific. of extratropical cyclones. E 19, 338-356. C. W. Newton and E. O. temperature gradients and advections from single- Holopainen, Eds., Amer. Meteor. Soc., 27- 45. Neiman, P. J., and M. A. Shapiro, 1989: RetrievingWea. Forecasting, horizontal 4, oceanic storms using thermal wind: a favorable Truitt, J. B., 1994: West-coast analysis of approachingFirst WSR- station wind profiler observations. 88D Users Conf., 222-233. coincidence for WSR-88D use. Postprints, J. Appl. Meteor. Norman, OK, 74-82. [Available from Parish, T. R., 1982: Barrier winds along the Sierra Nevada NOAA/WSR-88D Operational Support Facility, 1313 Mountains. , 21, 925-930. Halley Circle, Norman, OK 73069.] of extratropical cyclone development using a scale- Extratropical Parsons, K. E., and P. E. Mon.Smith, Wea. 2004: Rev., An investigation Uccellini,Cyclones: L. W., The 1990: Erik Processes Palmén Memorialcontributing Volume, to the rapid development of extratropical cyclones. separation technique.Motion and Motion 132, Systems. 956-974. C. W. Weather Analysis and Forecasting, Newton and E. O. Holopainen, Eds., Amer. Meteor. Soc., Petterssen, S., 1956: Vol. 1, 81-105. McGraw-Hill, 428 Images in Weather pp. Young,Forecasting, M. V., and J. R. Grant, 1995: Interpreting features associated with baroclinic troughs. M. J. Bader, G. S. Forbes, J. R. Grant, R. B. E. Lilley, and A. J. Waters, Eds., Cambridge University Acknowledgments Press, 287-301.

The author wishes to thank the two reviewers for the National Weather Digest for their constructive and highly detailed comments about the manuscript. They provided numerous needed recommendations. They are Dr. Jerome Patoux, Professor of Meteorology at the University of Washington, Seattle, and Dr. Scott Rochette, Associate Professor of Meteorology at the State University of New York (SUNY) at Brockport.

Dr. David Schultz, National Severe Storms Laboratory, provided both encouragement and an extra iteration during a prior review process that helped tighten the science and the text. Conversations with John Weaver, Research Meteorologist at NOAA, and Dan Bikos, Research Coordinator for the Cooperative Institute for Research in the Atmosphere, helped the cyclogenesis presentation. Tom Ainsworth, Meteorologist In Charge at WFO Juneau, provided suggestions to help readability as well as operational application. Gary Hufford, NWS Alaska Regional Scientist, and Audrey Rubel, Physical Scientist, provided a joint manuscript review and several of their recommendations have been applied.

Sarah Olsen served as computer graphics illustrator. Axel Graumann, meteorologist and Head of the Satellite Services Group within the National Climatic Data Center, provided the GOES IR image. Paul Shannon, then Lead Forecaster and currently the Information Technology Officer in Juneau, wrote the WSR-88D VWP script. Carl Dierking, Science and Operations Officer in Juneau, wrote the application used to log forecaster notes within the AWIPS environment, and provided a detailed review with several changes that have been applied.

Gary Ellrod, Chief Editor of National Weather Digest, encouraged elaboration on the possible satellite application of the method. He and Kermit Keeter, Technical Editor of National Weather Digest, provided finishing reviews that added to the text’s clarity.

Volume 32 Number 2 ~ December 2008 165