12
1040 VOLUME 130 MONTHLY WEATHER REVIEW q 2002 American Meteorological Society Case Study of a Coastal Jet at Spitsbergen—Comparison of SAR- and Model-Estimated Wind ANNE DAGRUN SANDVIK University of Bergen, Bergen, Norway BIRGITTE RUGAARD FUREVIK Nansen Environmental and Remote Sensing Center, Bergen, Norway (Manuscript received 20 March 2001, in final form 11 September 2001) ABSTRACT A combination of in situ ship measurements, synthetic aperture radar (SAR) imagery, and high-resolution numerical modeling was used to investigate a mesoscale coastal jet radiating out from Hinlopenstretet, Norway, on 14 August 1996. In the meteorological analysis, a light breeze and high static stability were found upstream of Svalbard. By studying results of numerical simulations in relation to the topography, upstream stagnation, flow splitting, and downstream jets and wakes were identified. In the Hinlopenstretet area this flow pattern was confirmed by in situ ship measurements and the SAR-estimated wind. Through investigations of sensitivity simulations, it was found that the low-level jet mainly was a result of stratified flow around dynamically steep, isolated topography, while the channeling effect of Hinlopenstretet had a minor influence. In the core of the jet the velocity was increased by a factor of about 3 compared to the upstream velocity. 1. Introduction A coastal jet was indicated when high backscattering from a limited sea surface area was seen on the satellite- based synthetic aperture radar (SAR) imagery acquired over Hinlopenstretet, Norway, between Spitsbergen and Nordaustlandet (Figs. 1 and 3), at 1149 UTC 14 August 1996. SAR images are snapshots of the surface roughness. For open water the surface roughness changes according to the wind stress, and areas of spatially inhomogeneous wind become visible in the radar images. Some features in an SAR scenes can also be related to oceanic phe- nomena like the long ocean surface wave field. How- ever, these are of second order and can often be ne- glected (Alpers et al. 1998). By using a wind scatter- ometer model, the Institut Franc ¸ais de Recherche pour l’Exploitation de la Mer (IFREMER) C-band model (CMOD-IFR2; Quilfen et al. 1998), the SAR measure- ments were converted to wind speed 10 m above the sea surface, assuming a neutrally stratified atmosphere and that the wind direction was known from ‘‘wind streaks’’ in the SAR image. From the SAR-estimated wind field a maximum ve- locity of a near gale was found in the indicated jet, while Corresponding author address: Anne Dagrun Sandvik, University of Bergen, Allegaten 70, N-5007 Bergen, Norway. E-mail: [email protected] a light breeze was found in the surroundings. This rel- atively high velocity exiting Hinlopenstretet was con- firmed by wind measurements from the research vessel (R/V) Ha ˚kon Mosby of the University of Bergen during a field experiment carried out north of Svalbard to pro- vide in situ data to be used together with the SAR images (Furevik et al. 2001, hereafter FJS). The meteorological analysis from the European Centre for Medium-Range Weather Forecasts (ECMWF) and the synoptic weather maps from the Norwegian Meteorological Institute (DNMI), show a relatively stationary synoptic situation, with light easterly winds over the island group. The local forecast from DNMI called for easterly breeze with no suggestion of gale, an underestimate that makes a large difference for planning of activity in the affected area. Mesoscale wind variations in the Arctic are important to document, not only for the obvious reason of weather forecasting. Wind is an important driving force for ocean circulation and also for mixing processes, which again have a significant impact on biological processes. In addition an improved understanding of mesoscale pro- cesses, especially in the Arctic, may be turned to account in global modeling of weather and climate. Hinlopenstretet is surrounded by complex topogra- phy, with mountains ranging from the largest scale, rep- resented by Spitsbergen, to the smallest hills (Fig. 1). The mountain slopes are in places very steep and the highest mountain on the island group is Newtontoppen Unauthenticated | Downloaded 04/20/22 07:43 AM UTC

Case Study of a Coastal Jet at Spitsbergen—Comparison of

  • Upload
    others

  • View
    1

  • Download
    0

Embed Size (px)

Citation preview

Page 1: Case Study of a Coastal Jet at Spitsbergen—Comparison of

1040 VOLUME 130M O N T H L Y W E A T H E R R E V I E W

q 2002 American Meteorological Society

Case Study of a Coastal Jet at Spitsbergen—Comparison of SAR- andModel-Estimated Wind

ANNE DAGRUN SANDVIK

University of Bergen, Bergen, Norway

BIRGITTE RUGAARD FUREVIK

Nansen Environmental and Remote Sensing Center, Bergen, Norway

(Manuscript received 20 March 2001, in final form 11 September 2001)

ABSTRACT

A combination of in situ ship measurements, synthetic aperture radar (SAR) imagery, and high-resolutionnumerical modeling was used to investigate a mesoscale coastal jet radiating out from Hinlopenstretet, Norway,on 14 August 1996. In the meteorological analysis, a light breeze and high static stability were found upstreamof Svalbard. By studying results of numerical simulations in relation to the topography, upstream stagnation,flow splitting, and downstream jets and wakes were identified. In the Hinlopenstretet area this flow pattern wasconfirmed by in situ ship measurements and the SAR-estimated wind. Through investigations of sensitivitysimulations, it was found that the low-level jet mainly was a result of stratified flow around dynamically steep,isolated topography, while the channeling effect of Hinlopenstretet had a minor influence. In the core of the jetthe velocity was increased by a factor of about 3 compared to the upstream velocity.

1. Introduction

A coastal jet was indicated when high backscatteringfrom a limited sea surface area was seen on the satellite-based synthetic aperture radar (SAR) imagery acquiredover Hinlopenstretet, Norway, between Spitsbergen andNordaustlandet (Figs. 1 and 3), at 1149 UTC 14 August1996.

SAR images are snapshots of the surface roughness.For open water the surface roughness changes accordingto the wind stress, and areas of spatially inhomogeneouswind become visible in the radar images. Some featuresin an SAR scenes can also be related to oceanic phe-nomena like the long ocean surface wave field. How-ever, these are of second order and can often be ne-glected (Alpers et al. 1998). By using a wind scatter-ometer model, the Institut Francais de Recherche pourl’Exploitation de la Mer (IFREMER) C-band model(CMOD-IFR2; Quilfen et al. 1998), the SAR measure-ments were converted to wind speed 10 m above thesea surface, assuming a neutrally stratified atmosphereand that the wind direction was known from ‘‘windstreaks’’ in the SAR image.

From the SAR-estimated wind field a maximum ve-locity of a near gale was found in the indicated jet, while

Corresponding author address: Anne Dagrun Sandvik, Universityof Bergen, Allegaten 70, N-5007 Bergen, Norway.E-mail: [email protected]

a light breeze was found in the surroundings. This rel-atively high velocity exiting Hinlopenstretet was con-firmed by wind measurements from the research vessel(R/V) Hakon Mosby of the University of Bergen duringa field experiment carried out north of Svalbard to pro-vide in situ data to be used together with the SAR images(Furevik et al. 2001, hereafter FJS). The meteorologicalanalysis from the European Centre for Medium-RangeWeather Forecasts (ECMWF) and the synoptic weathermaps from the Norwegian Meteorological Institute(DNMI), show a relatively stationary synoptic situation,with light easterly winds over the island group. The localforecast from DNMI called for easterly breeze with nosuggestion of gale, an underestimate that makes a largedifference for planning of activity in the affected area.

Mesoscale wind variations in the Arctic are importantto document, not only for the obvious reason of weatherforecasting. Wind is an important driving force for oceancirculation and also for mixing processes, which againhave a significant impact on biological processes. Inaddition an improved understanding of mesoscale pro-cesses, especially in the Arctic, may be turned to accountin global modeling of weather and climate.

Hinlopenstretet is surrounded by complex topogra-phy, with mountains ranging from the largest scale, rep-resented by Spitsbergen, to the smallest hills (Fig. 1).The mountain slopes are in places very steep and thehighest mountain on the island group is Newtontoppen

Unauthenticated | Downloaded 04/20/22 07:43 AM UTC

Page 2: Case Study of a Coastal Jet at Spitsbergen—Comparison of

APRIL 2002 1041S A N D V I K A N D F U R E V I K

FIG. 1. The geographical location of the 6-km (covering the Sval-bard archipelago) and 2-km domain, with names used in the text. Thesolid lines show the model topography with 200 m equidistance andthe hatched area to the north shows the Arctic ice edge used for thenumerical integrations.

(1717 m). With the normally high static stability at theselatitudes (Serreze et al. 1992), the airflow over the com-plex topography frequently becomes nonlinear and dif-ficult to prescribe (Skeie and Grønas 2000). There isalso limited confidence in the weather forecasts amongthe inhabitants of Svalbard, and people are aware ofsmall-scale atmospheric features due to the complex to-pography.

While mesoscale wind variations connected to ide-alized topography are well documented (Baines 1995;Durran 1990), less work has been done on real casestudies at high latitudes. Nevertheless, local areas ofincreased (jet) or decreased (wake) wind speed in coastalareas are well-known phenomena observed by remotesensing (Furevik and Korsbakken 2000; Pan and Smith,1999; Chelton et al. 2000a,b), reported for instance byGrønas (1997), and demonstrated by simulations inhigh-resolution numerical models, for example, with theHigh Resolution Limited Area Model (HIRLAM)weather prediction model (Gustafsson 1993; Kallen1996) using a horizontal grid length of 10 km. Thesephenomena are a result of the topography as well as theatmospheric conditions given by the static stability andthe large-scale wind field. An extensive study of jetsand wakes related to the Svalbard area may be foundin Skeie and Grønas (2000). The present Hinlopen jet

could be identified as a coastal jet due to a Bernoullieffect around the corner of Spitsbergen, or more pre-cisely as a tip jet according to the classification by Doyleand Shapiro (1999). It could also have been identifiedas a gap flow (Pan and Smith 1999), but as will beshown, the channeling effect through Hinlopenstretethas only minor influence on the jet.

To get a 3D picture of the Hinlopen jet and to extractmaximum information from the SAR images, it wasdecided to make high-resolution nested integrationswith the nonhydrostatic Pennsylvania State University–National Center for Atmospheric Research fifth-gener-ation Mesoscale Model (MM5; Dudhia 1993; Grell etal. 1994). Analyses from ECMWF were used for theinitialization and lateral boundary values. In this harshclimate, with limited human activities, there are ratherfew meteorological observations. This causes some un-certainty in the meteorological analysis, which furtherinfluences the quality of limited area weather forecasts.In addition to the quality of the meteorological analysis,it is also a fact that many of the Arctic weather phe-nomena in this area are on a scale too small to be re-solved by operational weather prediction models, likeHIRLAM or ECMWF.

The most obvious benefit from the numerical simu-lations is that they exemplify the value of increasedmodel resolution on the quality of local weather fore-casts in the area. Another motivation is that model re-sults can be used to extend vertically the essentiallysurface description afforded by the observations. In ad-dition sensitivity integrations, where large topographicalfeatures were eliminated, give the possibility to increas-ing our understanding of the cause of the jet.

In section 2 the observational basis for this study ispresented. The numerical model and the setup of theexperiments are described in section 3. Then the resultsare discussed in section 4.

2. Observations and methods

The two SAR scenes were acquired during an ice edgeexperiment north and northwest of Svalbard on 6–15August 1996 (Johannessen and Haugan 1997). The fieldexperiment was conducted by the R/V Hakon Mosby.This section starts with a discussion of the synopticsituation and the quality of the ECMWF analysis. Nexta short introduction to the SAR is given, and finally insitu ship measurements are discussed.

a. The ECMWF analysis and synoptic observations

During the period 14–16 August 1996 a weak easterly(about 3–5 m s21) flow across Svalbard (Fig. 2) wasfound near the surface in the ECMWF analysis. Up-stream of Svalbard, the analysis shows a very thin (100m) convective boundary layer and an inversion from100 to 800 m above the surface (not shown). This is atypical temperature profile when air masses flow from

Unauthenticated | Downloaded 04/20/22 07:43 AM UTC

Page 3: Case Study of a Coastal Jet at Spitsbergen—Comparison of

1042 VOLUME 130M O N T H L Y W E A T H E R R E V I E W

FIG. 2. The ECMWF analysis at 1200 UTC 14 Aug 1996. Thesolid lines and arrows show the height (20 m between isolines) ofand the wind velocity at the 1000-hPa surface, respectively.

a cold land area like sea ice and out over open water.An inversion develops over the ice, due to net radiationloss to the atmosphere. As the air flows out over theocean, a shallow convective layer develops. Sea surfacetemperatures (SSTs) of about 21.48C and air temper-atures of about 22.58C were found in the analysis eastof Svalbard. An ice map from DNMI shows that thearea was partly covered with ice.

As mentioned earlier, observations are rather sparsein this Arctic area, introducing uncertainty in the qualityof the analysis. It is also obvious that the coarse reso-lution (about 50 km horizontally and 6 h in time) is notadequate to include local effects of the topography. Theobserved and analyzed wind direction and speed werecompared for three synoptic weather stations: Hopen,Svalbard Airport (Longyearbyen), and Hornsund (forlocations see Fig. 1). At Hopen, which is a rather flatisland, there was acceptable agreement between boththe wind speed and direction. However, the wind speedis slightly underestimated (0–3 m s21) in the ECMWFanalysis. At Hornsund the wind directions agree well,but the wind speed in the analysis was much lower (2–5 m s21) than observed. At Longyearbyen there are alsosmall-scale features in the flow that are not captured inthe analysis. Both Hornsund and Longyearbyen are lo-cated downstream of significant topography. North ofNordaustlandet about 10 m s21 was recorded on boardthe Hakon Mosby, while 4.5 m s21 was found in theanalysis. To summarize, it seems like the analysis fromECMWF underestimated the wind speed in this area.This is in agreement with the wind speed from the hind-cast database (Eide et al. 1985; Reistad and Iden 1998)

from DNMI. The hindcast wind speed upstream of Nor-daustlandet is 4–8 m s21.

b. SAR

Figure 3 shows the wind speed estimated from twoSAR images acquired over Hinlopenstretet at 1149 UTC14 August 1996. There are great local variations, andthe wind speed varies from zero in the wakes to 14 m s21

in the core of the jet.The great advantage of the SAR images of the ocean

surface from the European Remote Sensing (ERS) sat-ellite is their high spatial resolution (30 m). SAR mea-surements can be used to estimate wind speed over openwater, and might be especially valuable in near-coastalareas where observational data often are too sparse togive a representative picture of the local wind field.

A system of short capillary and gravity waves is cre-ated instantly by the wind and behaves as an indicatorof the surface wind speed. The radar signal is reflectedby these waves, which are on a scale comparable to thewavelength (5 cm) of the signal. Empirical relationsbetween the wind speed (normalized to 10 m above thesurface) and the received reflected radar signal from thesurface have been established using buoy data or modeloutput. These relations are known as C-band models (inthe present work CMOD-IFR2 is used), and their ap-plications and accuracy for SAR are reported in, forexample, Vachon and Dobson (1996), Scoon et al.(1996), and Korsbakken et al. (1998). Comparisons withmodel output for a study along the Italian coast and theBaltic Sea have been reported in Alpers et al. (1998)and in Lehner et al. (1998), respectively. The SAR-estimated wind shows good results when compared toboth in situ measurements and model output. Roughlythe accuracy can be summarized to lay within 2 m s21.A similar result was reported for an Arctic area in FJS.However, more validation was recommended due to thelimited data available. The main disadvantage of SARimages is that they are ‘‘snapshots’’ with coarse timeresolution (long time between passages over the samearea).

c. Hakon Mosby data

In August 1996 the Hakon Mosby carried out a fieldexperiment north of Svalbard to provide in situ data. Atthe time of the SAR passage the Hakon Mosby was about40 km east of Hinlopenstretet. The ship track (shownin Fig. 3) then went west and into the jet about 1 hbefore midnight (11 h after the SAR image was ac-quired). The wind measurements from the Hakon Mosbyshow increasing wind speed as the ship moved westwardagainst Hinlopenstretet on 14 August.

3. MM5 and setup of the experiment

The nonhydrostatic MM5 was used for the simula-tions of the Hinlopen jet. The model has been exten-

Unauthenticated | Downloaded 04/20/22 07:43 AM UTC

Page 4: Case Study of a Coastal Jet at Spitsbergen—Comparison of

APRIL 2002 1043S A N D V I K A N D F U R E V I K

FIG. 3. The wind speed (10 m above the ground, pixel size 400 m 3 400 m) estimated from thetwo ERS-2 SAR scenes acquired over Hinlopenstretet at 1149 UTC, 14 Aug 1996. Wind direction is101.58 (northern scene) and 163.98 (southern scene). The ship track of the R/V Hakon Mosby is shownas the winding line; stars marks the position every 2 h from 1139 UTC to 0539 UTC next day. LinesAB and CD are profiles to be used in Fig. 9. These SAR scenes are taken from FJS.

sively used for research purposes and is still under de-velopment. The MM5 accommodates multinested do-mains and various physical parameterizations. For a de-tailed description see Dudhia (1993) and Grell et al.(1994).

For the present case study the model was configuredwith two domains with horizontal grid resolution of 6and 2 km. The outermost nest covers the Svalbard ar-chipelago, while the innermost nest covers Hinlopen-stretet as seen in Fig. 1.

The topographical data were taken from global U.S.Geological Survey (USGS) datasets with horizontal res-olution of 300 (0.925 km in the N–S direction). Thisregular latitude–longitude datasets are then interpolatedto the stereographic mesoscale domain selected for thesesimulations.

The vertical coordinate, spressure, is defined as s 5(p0 2 pt)/(ps0 2 pt), where p0 is the initial preference

state pressure, ps0 is the initial surface pressure, and pt

is a specified constant top pressure. Then, s becomesindependent of time. The simulations were performedwith 33 vertical levels, where the lowermost level forthe main prognostic variables was set at about 16 mabove the ground, and the top level was set to 100 hPa(15 km).

In the present study the following physical parame-terizations were activated: a nonlocal turbulence param-eterization (Hong and Pan 1996), an advanced soil mod-el (to include ice-covered ocean), a radiation parame-terization scheme taking into account the effect ofclouds on shortwave and downward longwave radiation(Benjamin 1983), an upper radiative boundary condition(Grell et al. 1994), and mixed phase microphysics withcloud water, rain, snow, ice, graupel, number of icecrystal, number of snowflakes, and number of graupelas prognostic variables (Reisner et al. 1998). At the

Unauthenticated | Downloaded 04/20/22 07:43 AM UTC

Page 5: Case Study of a Coastal Jet at Spitsbergen—Comparison of

1044 VOLUME 130M O N T H L Y W E A T H E R R E V I E W

FIG. 4. Model-predicted horizontal wind pattern (velocity and wind arrows), 16 m above thesurface, at 1200 UTC 14 Aug 1996, in the 6-km integration. The lines marked A and B are usedfor the vertical cross sections in Fig. 6. The tick marks on the axes show gridpoint values (6 kmbetween each tick mark).

selected model resolutions, convection is assumed to besolved explicitly.

The model was initialized with analyses of upper airand surface data from the ECMWF, and lateral boundaryvalues were included every 6 h during the 18 h of in-tegration starting at 0000 UTC 14 August 1996. Sea icewas initiated where the SST was less than 21.98C. The6-km domain was run concurrently with the 2-km do-main (two-way nesting), and the time steps were 18 and6 s, respectively. Relaxation zone lateral boundary con-ditions were used for the outermost domain, while theinnermost domain was updated every coarse-mesh timestep.

4. Results and discussion

The results from the numerical integrations on the 6-and 2-km model domains are discussed and comparedwith the SAR-derived wind speed and observationaldata from the Hakon Mosby. Furthermore, details in the2-km simulation are discussed, and finally the sensitivitysimulations are discussed in section 4d.

a. Results from the numerical model

For the given weak easterly flow, the static stabilityupstream of Svalbard (as described in section 2a) willhighly influence the local flow pattern. The effectivemountain height h 5 (NhM)/U, where N is the Brunt–Vaisala frequency, hM is the maximum height of themountain, and U is the horizontal velocity, will be veryhigh (roughly estimated to be 6), resulting in a flowregime with flow splitting and upstream stagnation asreported by Pierrehumbert and Wyman (1985), Smith(1989), Smolarkievicz and Rotunno (1990), and Olafs-son and Bougeault (1996) for idealized flow. There willbe increased velocities on both sides of the mountaindue to a narrowing of the streamlines. The Rossby num-ber, Ro 5 U/ fLx, where f is the Coriolis parameter andLx is the mountain width in the x direction, will be ona scale of (Ro ; 1) where the flow is clearly affected,but not dominated, by the rotation of the earth. On sym-metric mountains this will result in a stronger windspeed on the left-hand side of the mountain than on theright-hand side.

Unauthenticated | Downloaded 04/20/22 07:43 AM UTC

Page 6: Case Study of a Coastal Jet at Spitsbergen—Comparison of

APRIL 2002 1045S A N D V I K A N D F U R E V I K

FIG. 5. Trajectories from 10 to 118 h, at model level 3 (about100 m above the surface).

FIG. 6. (a) Vertical cross section of the horizontal wind speed, alongline A (432 km) in Fig. 4. Isolines are shown for each 2 m s21, from0 to 14 m s21. The positions used for the vertical soundings presentedin Fig. 7 are marked with I and II. Left on the figure is west. (b)Same as in (a) but along line B in Fig. 4.

To understand the reason for the Hinlopen jet, it isappropriate to first evaluate the main features in the 6-km flow pattern. The velocity and wind arrows in thelowermost model level (16 m above ground) at 1200UTC are presented in Fig. 4. Stagnation points and flowsplitting are seen upstream of the main topographicalfeatures, for example, east of Nordaustlandet (NL in Fig.4). Increased wind velocities (both on the right- andleft-hand side) are seen at every ‘‘corner’’ in the to-pography and wakes are seen downstream. The mostpronounced speedup is found out of Hinlopenstretet,where the maximum velocity has increased to about 13m s21. This velocity is enhanced by a factor of about3 relative to the upstream undisturbed flow (east of Nor-daustlandet). From trajectories (10 to 118 h; Fig. 5)made about 100 m above the ground, it is clearly seenthat the low-level flow in the present case goes aroundthe mountains.

A vertical cross section from west to east along lineA in Fig. 4, through the position of maximum velocity,is shown in Fig. 6a. The small peak to the left of thejet is the northern tip of Spitsbergen, and the mountainsto the right of the jet are Nordaustlandet. As seen, thisis a low-level jet, which sticks close to Spitsbergen. Adistinct signature of the jet might be seen to about 800m above surface. A vertical cross section from SE toNW through Hinlopenstretet (line B in Fig. 4) is shownin Fig. 6b. As seen, there is a wave signature in the

velocity field. A further investigation showed that colddense air is forced through Hinlopenstretet, and it wedg-es under warmer air north of Spitsbergen.

A closer look into the upstream atmospheric condi-tions (location II in Fig. 6a) is shown with the dashedlines in Figs. 7a and 7b. The vertical temperature profileupstream of Nordaustlandet (dashed line in Fig. 7a)shows a nearly neutral static stability in the lowest 100m and an inversion from about 100 to 750 m above thesea surface. The difference between the sea and air tem-perature at this position is about 1.5 K, resulting in aheat flux from the ocean into the atmosphere.

A nearly uniform wind speed, ranging from 3 to 5m s21 is found at the upstream location (Fig. 7b). In thecore of the jet (location I in Fig. 6a) a wind speedmaximum of 15.5 m s21 is found between 100 and 150m above the ground. It then decreases rapidly withheight as the static stability is strengthened. At the

Unauthenticated | Downloaded 04/20/22 07:43 AM UTC

Page 7: Case Study of a Coastal Jet at Spitsbergen—Comparison of

1046 VOLUME 130M O N T H L Y W E A T H E R R E V I E W

FIG. 7. Vertical (a) temperature and (b) velocity profiles, upstrream (position II in Fig. 6) and in the core of the jet (position I in Fig. 6).The star indicates the single 6-km integration.

height of maximum horizontal velocity the wind speedis increased with a factor of more than 4 compared tothe upstream velocity.

A plot, similar to that in Fig. 4, from the 2-km in-tegration is shown in Fig. 8. Except for a slight increasein the maximum velocities, the 2-km simulation gavethe same results as the 6-km simulation. Figure 7b alsoshows that increasing the horizontal resolution to 2 kmonly seems to have a minor effect. The wind speedincreased about 0.5 m s21 at location I. The small dif-ference is mainly due to the two-way nesting, where theinnermost domain affects the outermost. The linemarked with a star in Fig. 7b shows that the single 6-km integration gave about 1.5 m s21 lower velocity thanthe comparative nested 6-km integration.

b. MM5 data compared with SAR and observationsfrom R/V Hakon Mosby

Based on the similarity between the MM5 wind fieldin Fig. 8 and the SAR-estimated wind shown in Fig. 3,it was concluded that there is good agreement betweenthe two totally different wind-estimating methods. Bothmethods gave a clearly defined jet exiting Hinlopen-stretet and a wake behind Nordaustlandet. An absolutemaximum of the jet was found to be 15.5 m s21 at modellevel 5 (116 m) at 1200 UTC. Maximum velocity in thelowermost level was found to be 13.8 m s21. In a scatterdiagram, the data from SAR versus MM5 were shownto have a slope of one, but with some scatter and a biasof 2 m s21 (FJS). It might also be seen from Figs. 3and 8 that along the main flow direction the MM5 winds

decrease more rapidly and turn left as they leave thegap of Hinlopenstretet.

The similarities and differences between the two windestimates might be exemplified by comparing the ve-locities along a track in the direction of the jet and acrossthe jet, as marked in Figs. 3 and 8. The wind speedalong track AB is presented in Fig. 9a. For comparisonwith the 2-km model results, a 2-km-averaged SARwind is used. As seen, there is good agreement betweenSAR- and MM5-estimated wind speeds at the locationwhere the MM5 jet has its maximum. However, thedifference is about 4 m s21 at location B, indicating thatthe model wind speed decreases too fast along the flowdirection. Along track CD (Fig. 9b) there is a high de-gree of similarity, but the model wind speed is 3–4 ms21 lower than the SAR wind. It might also be seen thatthe MM5 is shifted slightly westward in accordance withthe turning of the MM5 wind as it leaves the gap.

From the model-estimated wind (not shown) it wasfound that the jet is steady with maximum velocity, inthe lowermost level, varying less than 1.5 m s21 from0900 to 1500 UTC. Despite the stationarity in the coreof the jet, some smaller fluctuations might be seen withincreasing size with the distance from the core. Fromthe SAR-estimated wind small spatial signatures mayindicate small rapidly changing features and vortexshedding.

The wind speed measurements from the Hakon Mosbyalong the ship track (Fig. 3) are shown in Fig. 10a, wherezero time on the x axis is the time of the SAR passage.To compare this time series with the SAR snapshot, thespatial variation along the ship track in the SAR-re-

Unauthenticated | Downloaded 04/20/22 07:43 AM UTC

Page 8: Case Study of a Coastal Jet at Spitsbergen—Comparison of

APRIL 2002 1047S A N D V I K A N D F U R E V I K

FIG. 8. Same as in Fig. 4 but for the 2-km integration. Isolines are shown every 2 m s21, from 0to 12 m s21. Lines AB and CD are profiles shown in Fig. 9.

trieved wind speed images is shown as a time series inthe same figure. The same procedure has been used forthe model results in the lowermost level at 1200 UTC.The time ‘‘correct’’ model wind is shown by stars at 1-h intervals. The dotted (MM5 112 h) line should thenbe compared to the SAR data, while the stars should becompared to the R/V Hakon Mosby data.

To examine the validity of making a time series froma spatial series, the stationarity in the flow pattern shouldbe investigated. Spatial series from MM5 every hourfrom 0600 to 2400 UTC were made. The means of the18 spatial series are shown as a time series with thebold line in Fig. 10b. Their standard deviations areshown as the dotted lines. The forecast situation is sta-tionary during the first 6 h of the ship track, then thereare larger differences, especially around 17 h and from19 to 111 h, which are the times when the ship isentering and leaving the simulated wake.

As discussed in FJS, there is a high degree of cor-relation between the R/V Hakon Mosby and SAR data

(Fig. 10a). Despite the obvious differences, there alsoseems to be good agreement between SAR and MM5.The MM5 field is smoother than the SAR field. Despitethe differences in the wake the shape of the two curvesis similar. However, the model wind speed is about 4m s21 weaker than the SAR wind in the jet. The overallconclusion from these comparisons is that there is a highdegree of similarity between the wind field from the twoestimation methods, but the numerical model seems tounderestimate the velocity. The reason for this is notobvious. From the discussion of the quality of ECMWFdata in section 2a, it is believed that the main reason isfound in the analysis. As marked with the black dot inFig. 10, the ECMWF wind speed at the ship positionat 1200 UTC is only 4.5 m s21. It is also worth notingthat the downstream flow pattern is very sensitive to theupstream flow direction, as for instance shown in Grønasand Sandvik (1999). Another possible reason could bethat the surface friction is too strong in the model. Sincethe situation is relatively stationary it can hardly be due

Unauthenticated | Downloaded 04/20/22 07:43 AM UTC

Page 9: Case Study of a Coastal Jet at Spitsbergen—Comparison of

1048 VOLUME 130M O N T H L Y W E A T H E R R E V I E W

FIG. 9. (a) SAR- (solid line) and MM5- (dashed line) estimatedwinds along line AB in Fig. 3. (b) Same as in (a) but along line CDin Fig. 3.

FIG. 10. Wind measurements from the Hakon Mosby (HK ) (solid line shows 10-minmean values) along the HM track shown in Fig. 3. The dashed line shows the SAR-estimated wind along the same track and the dotted line shows model results from 1200UTC. The star show the model wind at the time and position of HM. (b) the solid lineis the mean velocity (1200–2400 UTC) along the ship track and the dotted lines showthe standard deviation. The SAR-estimated wind (dashed line) is included for comparison.

to a time lag. However, the model wind 3 h earlier and100 m higher is very close to the SAR wind (not shown).On the other hand, it could be the SAR-estimated windsthat are too high. However, the ship measurements in-dicate that they are not, at least not for the position ofthe ship at the time of satellite passage. Also the extent

of the jet is indicating higher velocity in the SAR jetthan in the model jet.

c. Sensitivity to resolution

Earlier studies in this area (Skeie and Grønas 2000)have shown that a horizontal grid length of 10 km de-scribes the main topographical features of Svalbard.However, the complexity in the topography also indi-cates that smaller-scale features might be important forthe flow.

Two separate integrations with horizontal resolutionsof 10 and 6 km were performed in addition to the nested6 ↔ 2 km integration, to illustrate the effect of increasedmodel resolution on the flow. Both the 10- and 6-kmintegration show the main features described in section4a, but compared to the nested 2-km integration theyshow clear signs of their coarser resolution. For ex-ample, the maximum velocity in the jet increases byabout 2.5 m s21 from 10 to 6 km. This is about the sameincrease found between the single 6-km integration andthe nested 2-km integration in section 4a (see Fig. 7b),which exemplifies the importance of model resolutionfor comparison with in situ and SAR data.

Unauthenticated | Downloaded 04/20/22 07:43 AM UTC

Page 10: Case Study of a Coastal Jet at Spitsbergen—Comparison of

APRIL 2002 1049S A N D V I K A N D F U R E V I K

FIG. 11. Same as in Fig. 4 but from the integration without Nor-daustlandet. The axis labels show gridpoint values (grid cell size is6 km).

FIG. 12. Same as in Fig. 4 but from the integration withoutSpitsbergen. The axis labels show gridpoint values (grid cell sizeis 6 km).

d. Sensitivity simulations with changing topography

Since the main features in the flow are seen in the 6-km simulation, the following sensitivity simulations aredone only with this resolution. Only minor differencesin the results were observed due to changing the numberof vertical layers. The following integrations we there-fore performed with 23 vertical layers.

As discussed in the previous section, the main flowfeatures are represented in the 10-km integration. There-fore, it is reasonable to assume that it is the largest (100km) topographical features that have the most influenceon the resulting flow pattern.

To investigate the effect of the main island Spitsber-gen, the island Nordaustlandet was removed from themodel topography. The resulting flow pattern, at 1200UTC, is shown in Fig. 11. There is still a significantincrease in the velocity due to Spitsbergen, but the max-imum velocity in the jet is about 2 m s21 lower. If wedefine the jet to be limited to the area where the velocitycompared to the background velocity is larger than 1.5,the cross-flow extent of the jet increases by about 20km (from 50 to 30 km). The horizontal extent is about220 km for both cases. Since this coastal jet was foundon the right-hand side of Spitsbergen, this is what isknown as a ‘‘tip jet’’ in the literature (Doyle and Shapiro1999). Results from a similar integration where Spits-bergen was removed from the model topography areseen in Fig. 12. Nordaustlandet is not able to create the

jet, but it increases the Hinlopen inflow velocity by 1–2 m s21, an increase that will influence the jet. Thechannel between Spitsbergen and Nordaustlandet, whenboth are represented in the model topography, alsoseems to have a certain effect on the formation and shapeof the jet.

5. Summary and concluding remarks

High-resolution numerical integrations with the non-hydrostatic MM5 were performed for an area coveringthe Svalbard archipelago on 14 August 1996. The modelresults show a coastal jet exiting Hinlopenstretet and awake downstream of Nordaustlandet. A similar jet anda similar wake were estimated from ERS-2 SAR mea-surements. Furthermore, the two phenomena were alsoconfirmed by observational data from the Hakon Mosby.

The signature in the velocity field might be seen fardownstream of Hinlopenstretet. In the 6-km integration(Fig. 4) the horizontal extent along the flow directioncan be estimated to about 2–300 km and the width to30–50 km. From the model data it was found that inthe core of the jet the wind speed was increased by afactor of about 3 compared to the upstream wind speed.The model data show a stationary situation, which ispreferable with respect to comparison of the SAR snap-shot data and the in situ measurements from the HakonMosby.

Unauthenticated | Downloaded 04/20/22 07:43 AM UTC

Page 11: Case Study of a Coastal Jet at Spitsbergen—Comparison of

1050 VOLUME 130M O N T H L Y W E A T H E R R E V I E W

Both the SAR estimations and the observations fromthe Hakon Mosby indicated that MM5 underestimatedthe wind speed in this particular case. There are reasonsto believe that this was due to an underestimation ofthe wind speeds in the ECMWF analysis that were usedfor initialization and the lateral boundaries. Neverthe-less, it should also be kept in mind that the velocitynear the surface will be affected by a number of param-eterizations, in particular the surface friction and thechoice of model resolution.

Through the simulations we have demonstrated theenormous effect that topography can have on the flowunder special atmospheric conditions. An increase inthe velocity by a factor of about 3 is rather high, andfor human activity it would have been of the greatestinterest to know whether a light breeze or a near galewas expected. The jet could hardly have been forecastwith the present operational tools, and this work illus-trates the value of high-resolution numerical models inpredicting detailed local forecasts of mountain flows.

The understanding of the development of the jet wasfurther improved through additional sensitivity simu-lations with major topographical displacement. It wasfound that the development of the jet mainly was a resultof flow around the mountains, rather than a channelingeffect, which had a minor, but not insignificant, effect.The authors have no documentation on the frequencyof such jets, but with a dominating easterly flow in thearea and a normally high static stability, such jets shouldbe relatively frequent.

The study also provides valuable insight into the useof a nonhydrostatic atmospheric model in the arctic cli-mate and special topography in the area.

Global weather and climate are sensitive to processesand interactions in the polar atmosphere (Houghton etal. 1995), and more effort should be spent on studiesto improve our knowledge of mesoscale wind variationsat high latitudes and the influence this will have onmesoscale ocean circulation. High-resolution numericalmodeling in combination with high-resolution SAR im-ages and in this context will be a powerful tool for thefuture.

Acknowledgments. The authors wish to acknowledgethe Atmospheric Mesoscale Group at the GeophysicalInstitute, University of Bergen, for useful discussions.We thank Ola M. Johannessen (NERSC) for providingthe experimental in situ data, and Idar Hessevik(GFIUB) for IT support locally. This work was sup-ported by the Norwegian Research Council, Projects122462/720 and 135972/410, and the Norwegian SpaceCenter, Project JOP.8.3.3.03.00.2 Supercomputing re-sources were made available by the Norwegian Re-search Council. Data are provided by European SpaceAgency under Contract AOT.N303.

REFERENCES

Alpers, W., U. Pahl, and G. Gross, 1998: Katabatic wind fields incoastal areas studied by ERS-1 synthetic aperture radar imagery

and numerical modelling. J. Geophys. Res., 103 (C4), 7875–7886.

Baines, P. G., 1995: Topographic Effects in Stratifield Flows. Cam-bridge Monogr. on Mechanics, Cambridge University Press, 482pp.

Benjamin, S. G., 1983: Some effects of surface heating and topog-raphy on the regional severe storm environment. Ph.D. thesis,The Pennsylvania State University, 265 pp.

Chelton, D. B., H. Freilich, and S. K. Esbensen, 2000a: Satelliteobservations of the wind jets off the Pacific coast of CentralAmerica. Part I: Case studies and statistical characteristics. Mon.Wea. Rev., 128, 1993–2018.

——, ——, and ——, 2000b: Satellite observations of the wind jetsoff the Pacific coast of Central America. Part II: Regional re-lationships and dynamical considerations. Mon. Wea. Rev., 128,2019–2043.

Doyle, J. D., and M. A. Shapiro, 1999: Flow response to large-scaletopography: The Greenland tip jet. Tellus, 51A, 728–748.

Dudhia, J., 1993: A nonhydrostatic version of the Penn State–NCARMesoscale Model: Validation tests and simulation of an Atlanticcyclone and cold front. Mon. Wea. Rev., 121, 1493–1513.

Durran, D. R., 1990: Mountain waves and downslope winds. At-mospheric Processes over Complex Terrain, Meteor. Monogr.,No. 45, Amer. Meteor. Soc., 59–82.

Eide, L. I., M. Reistad, and J. Guddal, 1985: Database av beregnedevind og bølgeparametre for Nordsjøen, Norskehavet og Baren-tshavet, hver 6. time for arene 1955–81. DNMI, Oslo/Bergen,Norway, 38 pp.

Furevik, B. R., and E. Korsbakken, 2000: Comparison of derivedwind speed from Synthetic Aperture Radar and Scatterometerduring the ERS Tandem Phase. IEEE Trans. Geosci. RemoteSens., 38, 1113–1122.

——, O. M. Johannessen, and A. D. Sandvik, 2001: SAR retrievedwinds in polar regions—Comparison with in situ data and at-mospheric model output. IEEE Trans. Geosci. Remote Sens., inpress.

Grell, G. A., J. Dudhia, and D. R. Stuffer, 1994: A description of thefifth-generation Penn State/NCAR Mesoscale Model (MM5).NCAR/TN-3891IA, National Center for Atmospheric Research,Boulder, CO, 122 pp.

Grønas, S., 1997: Mesoscale phenomena induced by mountains overSkandinavia and Spitsbergen. Proc. Workshop on Orography,Reading, United Kingdom, ECMWF, 165–182.

——, and A. D. Sandvik, 1999: Numerical simulations of local windsover steep orography in the storm over north Norway on October12, 1996. J. Geophys. Res., 104, (D8), 9107–9120.

Gustafsson, N., 1993: HIRLAM 2 final report. HIRLAM Tech. Rep.9, Swedish Meteorological and Hydrological Institute, Norrko-ping, Sweden, 126 pp.

Hong, S.-Y., and H.-L. Pan, 1996: Nonlocal boundary layer verticaldiffusion in a medium-range forecast model. Mon. Wea. Rev.,124, 2322–2339.

Houghton, J. T., L. G. Meira Filho, B. A. Callender, N. Harris, A.Kattenberg, and K. Maskell, Eds., 1995: Climate Change 1995:The Science of Climate Change. Cambridge University Press,572 pp.

Johannessen, O. M., and V. E. Haugan, 1997: Pilot Arctic OceanProject 1996: Data report on mesoscale physical processes inthe marginal ice zone in the Barents Sea. NERSC Special Rep.50, Nansen Environmental and Remote Sensing Center, 270 pp.

Kallen, E., 1996: HIRLAM documentation manual, system 2.5. Swed-ish Meteorological and Hydrological Institute, Norrkoping, Swe-den.

Korsbakken, E., J. A. Johannessen, and O. M. Johannessen, 1998:Coastal wind field retrievals from ERS synthetic aperture radarimages. J. Geophys. Res., 103, (C4), 7857–7874.

Lehner, S., J. Horstmann, W. Koch, and W. Rosenthal, 1998: Me-soscale wind measurements using recalibrated ERS SAR images.J. Geophys. Res., 103, (C4), 7847–7856.

Unauthenticated | Downloaded 04/20/22 07:43 AM UTC

Page 12: Case Study of a Coastal Jet at Spitsbergen—Comparison of

APRIL 2002 1051S A N D V I K A N D F U R E V I K

Olafsson, H., and P. Bougeault, 1996: Nonlinear flow past an ellipticmountain ridge. J. Atmos. Sci., 53, 2465–2489.

Pan, F., and R. B. Smith, 1999: Gap winds and wakes: SAR obser-vations and numerical simulations. J. Atmos. Sci., 56, 905–923.

Pierrehumbert, R. T., and B. Wyman, 1985: Upstream effects of me-soscale mountains. J. Atmos. Sci., 42, 977–1001.

Quilfen, Y., B. Chapron, T. Elfouhaily, K. Katsaros, and J. Tournadre,1998: Observation of tropical cyclones by high-resolution scat-terometry. J. Geophys. Res., 103, 7767–7786.

Reisner, J., R. J. Rasmussen, and R. T. Bruintjes, 1998: Explicit fore-casting of supercooled liquid water in winter storms using theMM5 mesoscale model. Quart. J. Roy. Meteor. Soc., 124B,1071–1107.

Reistad, M., and K. A. Iden, 1998: Updating, correction and evalu-ation of a hindcast data base of air pressure, winds and wavesfor the North Sea, Norwegian Sea and the Barents Sea. DNMIRes. Rep. 9, 42 pp. [Available from Norwegian MeteorologicalInstitute, Pb. 43, Blindern, N-0313 Oslo, Norway.]

Scoon, A., I. S. Robinson, and P. J. Meadows, 1996: Demonstrationof an improved calibration scheme for ERS-1 SAR imagery us-ing a scatterometer wind model. Int. J. Remote Sens., 17, 413–418.

Serreze, M., J. Kahl, and R. Schnell, 1992: Low-level temperatureinversions of the Eurasian arctic and comparisons with Sovietdrifting station data. J. Climate, 5, 615–629.

Skeie, P., and S. Grønas, 2000: Strongly stratified easterly flows acrossSpitsbergen. Tellus, 52A, 473–486.

Smith, R., 1989: Hydrostatic airflow over mountains. Advances inGeophysics, Vol. 31, Academic Press, 59–81.

Smolarkievicz, P. K., and R. Rotunno, 1990: Low Froude numberflow past three-dimensional obstacles. Part II: Upwind flow re-versal zone. J. Atmos. Sci., 47, 1498–1511.

Vachon, P. W., and F. W. Dobson, 1996: Validation of wind vectorretrieval from ERS-1 SAR images over the ocean. Global Atmos.Ocean Syst., 5, 177–187.

Unauthenticated | Downloaded 04/20/22 07:43 AM UTC