mopdf.com

ARTICLE IN PRESS28M.D. Powell et al. / Ocean Engineering 37 (2010) 26–36stations are NOAA National Data Buoy Center and University ofSouthern Mississippi moored buoys, coastal platforms from theCoastal-Automated Marine Network (C-MAN), National OceanService (NOS), Louisiana Universities Marine Consortium(LUMCON), the Louisiana State University (LSU) Wave-CurrentSurge Information System for Coastal Louisiana (WAVECIS), theGulf of Mexico Ocean Observing System-Regional Association(GCOOS-RA), and the Weeks Bay, AL Network of the NationalEstuarine Research Reserve System.Marine observations were standardized as described in Powellet al. (1996), using the Liu surface layer model (Liu et al., 1979) tocompute the 10 min mean wind at 10 m level. The only change tothe Liu model was to use the Large and Pond drag coefficientrelationship for wind speeds o34 m s 1 and hold the dragcoefficient constant at 2 10 3 for winds above hurricane forceconsistent with observations from GPS sondes in hurricanesTable 1Coastal, marine, and inland weather networks.NetworkNumber ofstationsAnemometerheights (m)Averagingtime ve CISGCOOS-RANWS New OrleansNOSUSADODNDBC Moored BuoysNational Estuarine ResearchReserve System (Weeks Bay, 6–104–11105–104–11Peak 1/15 minPeak 1/10 minPeak 1/10 min102221, Peak 3 sgust10Variable1–21010106–101021010(Powell et al., 2003). The maximum 1 min sustained wind speedwas then computed by multiplying the mean marine surface windspeed by a gust factor (see below).3.2. Land-based observation platformsObserving platforms over land (Table 2) include portablemesonet stations deployed by the Florida Coastal MonitoringProgram (FCMP, 5 towers), Texas Tech University (TTU, 3 towers),University of Louisiana at Monroe (ULM, 1 tower), Low-LevelWind Shear Alert System network surrounding New Orleans (3stations), a New Orleans Weather Forecast Office automatedstation on the Lake Ponchartrain Causeway, an agriculturalnetwork of 25 stations operated by Louisiana, Mississippi, andAlabama as part of the Louisiana Agriclimatic Information System(LAIS), and observations logged by Emergency Operations Centersat NASA Michoud, Pearl River, and Pascagoula. Conventionalweather stations included the Automated Surface ObservingSystem (ASOS, 60 stations), Aviation weather stations (METAR,22 stations), and 28 miscellaneous stations from theMeteorological Assimilation Data Ingest System (MADIS).Winds measured by land platforms (including coastal platforms for some offshore wind directions) are influenced byfriction associated with upstream terrain features. In these casesstandardization requires knowledge of the upstream terrainroughness. The FCMP and TTU observations contained estimatesof surface roughness determined from measurements of turbulence intensity. For the remaining stations, roughness (Zo) wasestimated for each wind direction octant with upstream influencebased on photographic documentation (Powell et al., 2004) orusing aerial photographs and satellite imagery available on theweb by applications such as Google Earth. Roughness estimateswere subjective, based on experience guided by qualitativedescriptions such as Weiringa (1992). H*Wind provides aninterface to allow the scientist to edit the roughness table andzero-plane displacement heights and immediately compare theupdated wind value to neighboring stations. H*Wind provides atool to export data for plotting in Google Earth so the scientist canvisualize roughness influences on the flow.The H*Wind objective analysis requires all observations toconform to a marine exposure. For land stations, the medianupstream octant station roughness estimated from aerial and siteTable 2Selected mesonet and supplemental coastal and inland observing platforms.StationLat (deg)Lon (deg)Anemometer heightSamplingFCMP Stennis T0FCMP Belle Chase T1FCMP Galliano T2FCMP Pascagoula T3FCMP Gulfport T5TTU Slidell SBC ClearTTU Stennis SBC WhiteTTU VacherieLLWAS #2LLWAS #8LLWAS #9Buras dataLake Ponchartrain Cswy.NASA Michaud EOCLA-MS-AL AgNet AgricolaMS-AL Weeks BayMS-AL Middle BayUSM Buoy 42067Jackson County EOC, PascagoulaPoplarville Pearl Riv. Cty .210411.3MaxMaxMaxMaxMaxMaxMaxMax89.5177.630.81 min1 min1 min1 min1 min1 min1 min1 mineveryeveryeveryeveryeveryeveryeveryevery15 min15 min15 min15 min15 min10 min10 min10 min10 min consecutiveMax 1 min3 s continuous record10 min mean every 30 minPeak 3 s gustPeak 3 s gust

ARTICLE IN PRESSM.D. Powell et al. / Ocean Engineering 37 (2010) 26–3629photography was approximately open (0.05 m) with about 25% ofthe octant station values 4 0.1 m and 10% 40.25 m. Observationswere first converted to a wind speed at a level within theboundary layer ( 250 m) where winds are assumed to beequivalent for different terrains under neutral stability using Eq.(1), and then estimated for open terrain using Eq. (2).Using the neutral stability logarithmic wind profile law and theratio of the wind speed at 250 m (U250) to the wind speed (Uza) atanemometer height (Za),U250 ¼ UzaLnðð250 DÞ Zo ÞLnððZa DÞ Zo Þð1Þwhere D is the zero-plane displacement height. In practice, D wasrarely used unless the exposure for a site was extremely poor. The250 m wind is then used to estimate the 10 m level mean windspeed (Uopen) for open terrain (Zo 0.03 m),Uopen ¼ U250Lnð10 :03ÞLnð250 :03Þð2ÞThe open terrain mean winds were increased 17% to convert toa marine exposure for use in the analysis, consistent with Vickeryet al. (2009). Vickery et al. (2009, Fig. 12), using a 600 m boundarylayer height and 20 km fetch, found a 0.83 ratio of the fullytransitioned mean flow over open terrain to that over the openocean, compared to 21% and 19% increase from open to marineexposure for ESDU (1984) and Simiu and Scanlan (1996),respectively.The maximum 1 min sustained wind speed over the timeperiod of the mean wind (usually 10 min) is then computed,U1marine ¼ Umarine Gm60;600ð3ÞFig. 3. GBVTD wind field analysis for the 1.0 km level from the Slidell WSR-88DDoppler radar for 1010 UTC on 29 August.where Gm60,600 is the marine exposure gust factor based on themethod described by Vickery and Skerlj (2005).For Umarine o34 m s 1, the gust factor depends on the meanwind speed:Gm60;600 ¼ 1:069 þ1:51 10 3 Umarineð4ÞAs winds increase 434 m s 1, the gust factor tends to level off.Gm60;600 ¼ 1:094ð5ÞThe marine-adjusted winds were then compared to actualmarine observations (e.g. GPS sondes, buoys, SFMR observations,etc.) using H*Wind’s graphical interactive quality control (QC)tools. Marine-adjusted land observations inconsistent withneighboring marine counterparts were ‘‘flagged’’ (removed fromconsideration). The nearest neighbor QC process ensures that theanalysis is consistent with marine observations in coastal,offshore, and lake locations. Once the marine wind field analysisis completed, portions of the wind field over land locations areconverted back to open terrain, as depicted in the analysisgraphics described later in the paper, matching the original openterrain adjusted values.3.3. Land-based and airborne radarFour radar-based wind measurement methods were used toestimate winds near the top of the boundary layer. An advantageof Doppler radar techniques is the ability to determine winds overa relatively large area, resolving the azimuthal variation in themaximum wind.Fig. 4. H*Wind screen grab of NOAA 43 Airborne Doppler radar winds adjusted tothe surface (red) for 1232 UTC. Also plotted are 10 min observations from GrandIsle C-MAN station (GDIL1) in black (1000–1224 UTC) and Belle Chase FCMP towerin orange (1218–1351 UTC). (For interpretation of the references to color in thisfigure legend, the reader is referred to the web version of this article.)(1) Observations from the Slidell and Mobile WSR-88D Dopplerradars were used to generate dual-Doppler analysis at for the500 and 1000 m levels.(2) The ground based velocity track display (GBVTD) techniqueLee et al., 1999) was used to generate wind fields at the 500and 1500 m levels on 1010 UTC on 29 August (Fig. 3). Anadvantage of the GBVTD is that only one radar is needed toproduce a wind field.(3) The tail Doppler radars aboard the NOAA P3 aircraft scan in afore-aft sampling pattern which enables a dual-Doppler

ARTICLE IN PRESS30M.D. Powell et al. / Ocean Engineering 37 (2010) 26–36analysis technique (Gamache et al., 1995) to determine windsat the 500 and 1000 m levels (e.g. 1232 UTC, Fig. 4).(4) A fourth method involved evaluation of winds close to theland-based radar sites using the velocity azimuth display(VAD) technique. This method generates a vertical profile ofthe horizontal wind based on data from 2–9 km from theradar.For each method, winds at these levels were then adjusted to thesurface empirically, based on their comparison to marine observations in the same storm-relative location. The resulting winds werethen evaluated against observations from other platforms to see howwell they fit with in-situ surface measurements. In locations wherethey did not agree, the surface measurements were given precedence and the radar observations were ‘‘flagged’’ so they wouldnot be used in the final objective analysis.3.4. Aircraft reconnaissance observationsFlight-level observations were available from the NOAA P3 andAir Force Reserves C-130 aircraft, typically at the 70 hPa level near3 km altitude. The NOAA P3 aircraft also carried the steppedfrequency microwave radiometer (SFMR), which measures windspeed based on the microwave emission from sea foam at fivedifferent frequencies. The SFMR resolves the radial location of thesurface wind maximum with far greater accuracy than GPS sondeswhich can easily miss the location of maximum winds. A detailedcalibration–validation of the SFMR was conducted in 2004 and2005 and involved comparisons to over 400 GPS sondes (seeUhlhorn and Black, 2003; Uhlhorn et al., 2007). The SFMR isconsidered to be a high-accuracy marine platform comparable to10 m discus buoys and GPS sondes.Comparisons between flight-level and SFMR maximum windson each radial flight leg were used to develop specific surfaceadjustment methods for flight-level observations on each day for27, 28, and 29 August. These methods (Fig. 5) determined theradius and aziumuth of the maximum surface wind relative to themaximum flight-level wind and also specified the radial variationof the winds. The radius of surface wind maximum is typicallylocated at about 85% of the flight-level radius of maximum windPowell et al. (2009). The ratio of the maximum surface wind to themaximum flight-level wind (the slant reduction factor) variedbetween 28 August and landfall with higher values on the 28th. Avertical reduction factor was determined on each day for thevariation of reduction factor with radius outside the vicinity of theeyewall. The methods were applied to the Air Force flight-levelwinds when the NOAA SFMR data were not available.GPS sondes are routinely launched by both NOAA and Air Forcereconnaissance aircraft. Two surface wind estimates are provided.A surface wind is computed based on the 0.5 s sampled motion ofthe sonde near 10 m (see Hock and Franklin, 1999). In extremewinds, turbulence below 200 m makes it difficult for the GPSsonde to ‘‘find’’ enough satellites to do the wind computation. AGPS sonde surface measurement (even with 10 s filtering) showshigh variability representative of the flow features it happens tobe falling through, and the semi-Lagrangian measurement isdifficult to relate to a conventional anemometer averaging time.An alternative surface wind estimate relating the GPS sondesurface wind to the mean wind over the lowest 150 m (WL150)was developed by Franklin et al. (2003). The WL150-determinedsurface wind tends to be less variable than the surface wind and isconsidered a high quality observation during the quality controlprocess. During landfall, SFMR and GPS sonde observations wereavailable over Lake Ponchartrain and just offshore MississippiSound.3.5. Satellite observationsWind measurements were available from the Sea Windsscatterometer aboard QuikScat (Quilfen et al., 2007) and fromFig. 5. Azimuthal variation (left) of the slant reduction factor in Katrina on 28 August, 2005 (top) and 29 August (bottom). Right panels show variation in the reductionfactor as a function radius scaled by the radius of the flight-level maximum wind speed.

ARTICLE IN PRESSM.D. Powell et al. / Ocean Engineering 37 (2010) 26–36tracking visible GOES imagery for cloud motions at low levels. TheQuikScat winds sometimes have direction errors associated withpoor first guesses but tend to be contaminated by rain and cloudat winds over about 30 m s 1 but otherwise help fill-in areasmissed by the aircraft and help to identify the extent of thedamaging wind field. When standardizing QuikScat winds forH*Wind, we first determine a time scale for the wind measurement attributed to a grid cell by dividing the grid cell dimensionby the wind speed. A gust factor is then applied (Powell et al.,1996) to estimate the highest 1 min wind over the time scaleattributed to the grid cell. Cloud drift winds were computed bythe University of Wisconsin-NOAA Cooperative Institute forMeteorological Satellite Studies (CIMSS) at pressure levels below70.0 hPa. Cloud drift winds were adjusted to the surface followingthe method of Dunion et al. (2002).4. The evolution of Katrina’s wind fieldTime series of minimum sea-level pressure (Fig. 7) andintensity (Fig. 6) show Katrina’s evolution from a strong tropicalstorm on entry into the Gulf of Mexico (0000 UTC, 26 August) tomaximum intensity (1200–1800 UTC 28 August), to landfalls inLouisiana and Mississippi (1200, 1500 UTC, 29 August), and finallyto a decaying system (1200 30 August). Intensity is defined by themaximum 1 min sustained surface wind anywhere in the storm ata particular time. Intensities in Fig. 6 are based on H*Windanalyses. The change in sea-level central surface pressure on Fig. 7provides an alternate assessment of intensity change and is31sampled more frequently than the maximum wind since over a 6 hperiod the aircraft may sample the location (azimuth on a radialflight leg) of maximum winds only once but sample the center ofthe eye several times. Based on pressure, the period of Katrina’smost rapid intensification was from 0600 to 0900 on 28 Augustand the period with the most rapid weakening was 1500–1800 29August H*Wind analyses for 0900 UTC for 28 August and the 1800UTC 29 August are based on time periods with the most rapidchanges, so the maximum wind estimates are not necessarily atthe analysis time but capture the peak measured intensity within3 h of that time.As Katrina emerged from the Florida peninsula into the Gulf ofMexico, the intensity ramped up slowly but steadily from about72 kts (35 m s 1) at 0900 UTC 26 August to 87 kts (48 m s 1) at2100 UTC 27 August, while the pressure fell from 98.5 to 95.0 hPa.During this time period Katrina’s ability to maintain healthyconvection effectively shielded it from dissipative effects of windshear (McTaggart-Cowan et al., 2007) and the extent of hurricaneforce and 50 kt wind doubled. From 2100 UTC 27 August to 1200UTC on 28 August (Fig. 8), a period of rapid intensity changecommenced with the passage of Katrina over relatively deeplayers of warm water associated with the Gulf of Mexico LoopCurrent and a warm-core ring feature to the west (Fig. 9).Upwelling and ocean mixing associated with strong winds in thevicinity of Katrina’s core transported relatively warm water to thesurface, effectively removing a brake to intensification (in theabsence of these high ocean heat content features, a hurricanewould normally transport cooler subsurface water to the surface,inhibiting surface enthalpy fluxes). While passing over theseocean features, Katrina reached her maximum intensity of 139 kts(72 m s 1) while pressure fell to 90.5 hPa. During theintensification period the radius of maximum surface wind(Rmax) contracted from 50 to 25 km, and the extent ofhurricane, 50 kt, and tropical storm-force winds continued toFig. 6. Time series of 3 h maximum sustained surface wind speed.Fig. 7. Time series of 3 h central sea-level surface pressure.Fig. 8. H*Wind analysis for Katrina’s entrance into the Gulf of Mexico at 1200 UTC,28 August 2005. Wind speed contours in kts. Box in upper left shows radial extent(nm) of hurricane, 50 kt, and tropical storm strength winds in each quadrant.

ARTICLE IN PRESS32M.D. Powell et al. / Ocean Engineering 37 (2010) 26–36Fig. 9. Pre-Katrina ocean heat content (from Shay, in press) depicting Katrina’strack and Saffir–Simpson scale relative to positions of the Loop Current (LC),Florida Current (FC), and warm core ring (WCR). Ocean heat content based onaltimetry data from Jason-1, Geosat Follow-On, and Envisat data.Fig. 10. As in Fig. 8 but for landfall at 1200 UTC, 29 August 2005.increase. There was no clear indication of an eyewall replacementcycle (Willoughby et al., 1982; Houze, 2007), rather a‘‘superintensity’’ process may have been acting, through whichthe eyewall gains fuel from mixing with the eye (McTaggartCowan et al., 2007; Persing and Montgomery, 2003). For the nextseveral hours Katrina’s wind field maintained this size andintensity with the central pressure remaining below 91.0 hPaand the extent of hurricane force, 50 kt, and tropical storm-forcewinds at 130, 220, and 370 km, respectively through 2100 on 28August.During the next 15 h while Katrina approached land, thepressure rose while Katrina moved over Gulf of Mexico shelfwaters containing relatively low ocean heat content. As Katrina’scenter was making landfall in Louisiana at 1200 UTC on 29 August(Fig. 10), a series of outer rainbands formed which actedeffectively as an outer eyewall (Fig. 11). The outer eyewall(60 km radius) became predominant and contained windsslightly higher than those in the inner eyewall. The outereyewall feature was not particularly well defined on radar, norby the flight-level winds on the reconnaissance aircraft. While thepressure rose and the winds in the eyewall decreased to 102 kts(53 m s 1) as a consequence of angular momentum conservation,the radius of maximum wind increased, and the extent ofhurricane-, 50 kt, and tropical storm-force winds increased to200, 300, and 400 km.Radar reflectivity images (Fig. 11) from three different radars(NOAA-43 research aircraft, Slidell, and Mobile National WeatherService WSR-88D) show conflicting depictions of secondary outereyewalls brightness temperatures compared to the MorphedIntegrated Microwave Imagery (MIMIC) 85–92 GHz signal of lowearth orbiting satellites (Wimmers and Velden, 2007). The radarstend to agree on major features (inner eyewall and a series of spiralbands to the north and northeast of the storm center) while theMIMIC product suggests a coherent outer eyewall adjacent to a moatregion with lower brightness temperatures. Outer eyewall featuresdepicted in radar and MIMIC often show large changes in reflectivityand brightness temperature between the outer band and ‘‘moat’’region between the inner and outer eyewall. These large magnitudesin the outer eyewall often produce an interpretation and expectationof dramatic outer wind maxima. In-situ data show a much moresubtle signal, typically only 1–2 m s 1 above winds in the vicinity.Examination of radial passes of flight-level and SFMR surface windmeasurements indicated that surface outer maxima were not alwayspresent when flight-level maxima were evident, and vice versa.Sometimes the surface maximum was associated with the innereyewall while the flight-level wind maximum was in the outereyewall, and vice versa. Furthermore, over a 4–6 h period, as theouter bands continually generated and propagated, the location ofthe maxima would shift such that the outer maximum could belocated at different radii and azimuths from pass to pass, consistentwith the radar depictions of multiple outer rainbands. Whenconducting an objective analysis of such data, the subtle outermaxima tend to be smoothed out leaving the inner maximu

Reconstruction of Hurricane Katrina’s wind fields for storm surge and wave hindcasting Mark D. Powella,, Shirley Murilloa, Peter Dodgea, Eric Uhlhorna, John Gamachea, Vince Cardoneb, Andrew Coxb, Sonia Oteroc, Nick Carrascoc, Bachir Annanec, Russell St. Fleurc a NOAA-AOML Hurricane Researc