MARINE BOUNDARY LAYER AND TURBULENT FLUXES OVER THE BALTIC SEA: MEASUREMENTS AND MODELLING SVEN-ERIK GRYNING1 and EKATERINA BATCHVAROVA1,2 1 Risø National Laboratory, DK-4000 Roskilde Denmark; 2 National Institute of Meteorology and Hydrology, Bulgarian Academy of Sciences, Sofia, Bulgaria
(Received in final form 27 July 2001)
Abstract. Two weeks of measurements of the boundary-layer height over a small island (Christiansø) in the Baltic Sea are discussed. The meteorological conditions are characterised by positive heat flux over the sea. The boundary-layer height was simulated with two models, a simple applied high-resolution (2 km × 2 km) model, and the operational numerical weather prediction model HIRLAM (grid resolution of 22.5 km × 22.5 km). For southwesterly winds it was found that a relatively large island (Bornholm) lying 20-km upwind of the measuring site influences the boundary-layer height. In this situation the high-resolution simple applied model reproduces the characteristics of the boundary-layer height over the measuring site. Richardson-number based methods using data from simulations with the HIRLAM model fail, most likely because the island and the water fetch to the measuring site are about the size of the grid resolution of the HIRLAM model and therefore poorly resolved. For northerly winds, the water fetch to the measuring site is about 100 km. Both models reproduce the characteristics of the height of the marine boundary layer. This suggests that the HIRLAM model adequately resolves a water fetch of 100 km with respect to predictions of the height of the marine boundary layer. Keywords: Baltic Sea, HIRLAM model, Marine boundary-layer height, Radiosonde measurements, Richardson method, Turbulent fluxes of sensible and latent heat.
1. Introduction The wind climate of coastal areas often shows large differences over land and sea. The condition over land has received considerable attention while that over the sea surface has been studied much less. Hsu (1988) and Garratt (1990) give an overview of marine boundary-layer models and measurements. With offshore flow of cold air over warmer water, a convectively driven internal boundary layer forms in response to the upward buoyancy flux over the water. This layer thickens in the downstream direction. Most of the models of the growth of the internal boundarylayer in the coastal zone have been developed and validated for the unstable internal boundary layer over land (Källstrand and Smedman, 1997; Melas and Kambezidis, 1992: Batchvarova and Gryning, 1998, Batchvarova et al., 1999). However, they should in principle also apply over the sea in the coastal zone when the sea is warmer than the land. This topic has received little attention, and will be considered in this paper. Attempts to describe the growth of the stable boundary layer over the sea have been made by Hsu (1983) and Mulhearn (1981), both based Boundary-Layer Meteorology 103: 29–47, 2002. © 2002 Kluwer Academic Publishers. Printed in the Netherlands.
30
SVEN-ERIK GRYNING AND EKATERINA BATCHVAROVA
on dimensional arguments. The Mulhearn (1981) formulation was reproduced by Gryning and Joffre (1987) for the Øresund experiment (Gryning, 1985). Melas (1998) proposed a model based on an integrated heat-flux length scale suggested by Stull (1983), and Garratt (1987) studied the stable boundary-layer growth by use of a mesoscale numerical model. Hsu (1984) dealt with the formation of the unstable boundary layer over the sea in the coastal area. Vihma and Brümmer (2002) discuss the boundary-layer development and modifications to the flow field at the ice edge zone in the Baltic Sea. The structure and height of the coastal boundary layer over the sea surface is important for several reasons. The water that evaporates from the sea surface into the atmosphere is dispersed vertically through the action of turbulence and becomes mixed throughout the atmospheric boundary layer. Because the top of the boundary layer to a high degree acts as a lid, its depth is one of the parameters controlling the water content in the air over the sea surface, and therefore has a feedback on the evaporation from the water surface. It is also one of the fundamental parameters used to characterise the structure of the boundary layer. The marine boundary-layer depth influences the formation of clouds and ship cloud trails (Dorman, 1994), and also acts as a trapping layer for the propagation of radar and microwave signals thus creating radar ducts (Brooks et al., 1999). Anomalous propagation in weather radar reflectivity patterns is associated with the marine boundary-layer top. Failure to account for the anomalous propagation echoes in the estimation of the rainfall leads to erroneous results (Moszkowicz et al., 1994). As a part of a Pilot study on Evaporation and Precipitation over the Baltic Sea (PEP-in-BALTEX) measurements were carried out on a small island in the southern part of the Baltic Sea. The measurements comprise fluxes of sensible and latent heat. Data from an 8-month period formed the basis for a study by Rutgersson et al. (2001) on the parameterisations of latent and sensible heat fluxes over the sea. In the present paper we concentrate on the formation and characteristics of the boundary layer and model the marine boundary-layer height. The study is based on measurements from a two-week campaign when the programme was intensified by radiosoundings. The development of the boundary layer was inferred from the air temperature and humidity radiosonde profiles. The meteorological conditions were characterised by heat flux from the sea to the atmosphere, creating an unstable boundary layer over the sea.
2. Site and Measurements The Baltic Sea covers about 390,000 km2 extending from sub-arctic to temperate climate zones, and is elongated in the north-south direction being roughly 1300 km long and 300 km wide. The southern part of the Baltic Sea is called Baltic Proper, and is connected to the open sea through the Danish Straits.
MARINE BOUNDARY LAYER AND TURBULENT FLUXES OVER THE BALTIC SEA
31
2.1. S ITE The measuring activities were concentrated at a cluster of small granite islands in the Baltic Sea known collectively as Ertholmene; Figure 1 shows a map of the Baltic Proper with the position of Ertholmene marked with a cross. For wind directions in the sector 190◦ to 270◦ Ertholmene lies about 20 km downwind of Bornholm, Figure 2 (left panel). In the sector 270◦ to 45◦ the water fetch to the Swedish coast is about 100 km. The biggest island, Christiansø, is a 0.2 km2 large, beautifully preserved 17th century island fortress, Figure 2 (right panel), and forms, together with its smaller, 0.04 km2 , sister-island, Frederiksø, a natural harbour. The islands are cultural and environmental protected areas; Græsholm, northwest of Christiansø, is a wildlife refuge. Ertholmene is rather flat, although Christiansø reaches 20 metres in its central parts. Commonly the group of islands is also named Christiansø, and this practise will be used here. 2.2. T URBULENCE MEASUREMENTS Long term measurements (April 1998 through December 1999) of atmospheric turbulent fluxes of sensible heat, latent heat and momentum were carried out at an 8-m mast placed on a small granite island 1–2 m above sea level with an open sector to the sea of 120◦ to 300◦ through south, Figure 2. The measurements of heat and momentum fluxes were performed with a Kaijo-Denki DAT/TR-61B three-dimensional sonic anemometer at a height of 7 m on the mast. The sensible heat flux is derived using the air temperature from the sonic (virtual sound temperature). The latent heat flux is based on measurements of the fluctuations of water vapour content in the air performed with an open path infrared optical hygrometer (OPHIR) mounted 0.5 m above the sonic anemometer, Figure 3. The signals from the instruments were sampled with a frequency of 10 Hz. All mean values, variances and covariances are derived as 30-min averages with software developed at Risø (Risø National Laboratory). All measurements including the raw data have been transferred to Risø via the Internet and stored for subsequent analysis. Synoptic data were obtained near the lighthouse at Christiansø, a few hundred metres from the meteorological mast. 2.3. V ERTICAL PROFILES OF TEMPERATURE AND HUMIDITY During an intensive observation period from 24 October to 5 November 1998 the measurement programme was extended with a total of 24 radiosoundings at Christiansø. The soundings were performed by an AIR system (Atmospheric Instrumentation Research Inc.) with radiosondes of type IS-5A Intellisonde, measuring temperature, humidity and pressure (can be converted to height) with data sampling every two seconds. The soundings were performed daily at noon with an ascent velocity of about 1–3 m s−1 ; most of the sondes were launched under difficult conditions due to very strong winds. During the period from 31 October
32
SVEN-ERIK GRYNING AND EKATERINA BATCHVAROVA
Figure 1. Map of the Baltic Proper with land surfaces and islands dotted. A cross shows the location of Christiansø (Ertholmene) east of Bornholm. The figure covers the model domain for the simple applied model. The co-ordinate system refers to UTM34.
to 2 November the radiosounding programme was intensified to 4 to 6 soundings per day. The depth of the boundary layer was subjectively estimated from the soundings, based mainly on the profile of the potential temperature, and taken as the height where the potential temperature starts to increase, simultaneously considering the humidity profile. Figure 4 shows examples of potential temperature profiles from radiosoundings at 0900, 1500 and 2400 GMT on November 1, 1998. It can be seen that the potential temperature is near constant as a function of height above the sea, and up to typically 500 m, above which it starts to increase marking the top of the boundary layer. The bullet depicts the subjectively estimated top of the boundary layer.
MARINE BOUNDARY LAYER AND TURBULENT FLUXES OVER THE BALTIC SEA
33
Figure 2. The left panel shows the island of Bornholm. Ertholmene is the group of islands northeast of Bornholm that can be seen in the upper right corner of the frame and it is marked by a cross. Right panel is a map of Ertholmene with the position of the meteorology mast indicated by a cross. The co-ordinate systems refer to UTM34.
Figure 3. The meteorological mast. The picture is taken looking towards southwest.
34
SVEN-ERIK GRYNING AND EKATERINA BATCHVAROVA
Figure 4. Examples of radiosonde profiles of potential temperature on November 1, 1998. Bullets indicate the subjectively estimated boundary-layer heights.
3. Models The height of the boundary layer during the period with intensive measurements was modelled by a simple applied model, and by data from the operational numerical weather prediction model HIRLAM. 3.1. HIRLAM The HIgh Resolution Limited Area Model HIRLAM is a complete model system for operational weather forecasts maintained by national meteorological services in several countries, and covers northern Europe. The baseline forecast model is a hydrostatic, semi-implicit limited area Eulerian model (Källén, 1996), and is based on the primitive equations with temperature, pressure, humidity and horizontal wind velocity components as prognostic variables. Operationally, different local versions of the HIRLAM model are used, and in this study we use HIRLAM data provided by the Swedish Meteorological and Hydrological Institute. Sea surface temperature and ice cover measurements, in combination with satellite data, are used as surface boundary conditions for the Baltic Sea area. Outside the Baltic area, analysed data from ECMWF (European Center for Medium range Weather Forecast) are applied. At the lateral boundaries the model is forced with operational analyses from the global models at the ECMWF. The operational HIRLAM forecast, with a forecast period of 6 to 11 h, is used in this study. The horizontal grid resolution is 22.5 km × 22.5 km and there are 31 vertical levels. Output from the simulations with the HIRLAM model consists
MARINE BOUNDARY LAYER AND TURBULENT FLUXES OVER THE BALTIC SEA
35
of hourly surface values of the fluxes of latent and sensible heat in each gridpoint and profiles of wind (u and v components), temperature and humidity as a function of the geopotential height (given at the approximate levels 30, 150, 350, 600, 950, 1300, 1750, 2200, 2650 . . . metres). The data from the HIRLAM simulations that we apply in this study were also used in Rutgersson et al. (2001) where further details of the model and the simulations can be found. The height of the boundary layer does not form a part of the output from the HIRLAM model, but has to be estimated from the available data. In this study we applied and compared two methods to extract the boundary-layer height from the HIRLAM output data; both are based on a bulk Richardson-number approach. For both methods the boundary-layer height is defined as the height where the bulk Richardson number reaches a critical value, typically 0.25. The two methods differ in their definition of the bulk Richardson number. In the approach described in Sørensen (1998) the bulk Richardson number for the layer between the surface and the height z above the surface is given by the following expression RiB =
gz(θ(z) − θ(s)) . θ(s)(u(z)2 + v(z)2 )
(1)
The quantities θ(s) and θ(z) are the potential virtual temperatures at the surface and height z, respectively, u(z) and v(z) are the horizontal wind components at height z, and g is acceleration due to gravity. The height of the boundary layer is given by the smallest height z at which the bulk Richardson number takes a prescribed value. Sørensen (1998) reports that the critical value of the bulk Richardson number is in the range 0.15–0.35 for HIRLAM data, and recommends a value of 0.25 for operational use. Vogelezang and Holtslag (1996) suggest a Richardson number where the wind is defined with respect to the lowest model level, and a term that accounts for surface friction has been added. The modified Richardson number reads RiB =
gz(θ(z) − θ(s)) , θ(s)[(u(z) − u(s))2 + (v(z) − v(s))2 + bu2∗ ]
(2)
where b is a parameterisation constant, recommended by Vogelezang and Holtslag (1996) to be taken as 100. The critical Richardson number is taken as 0.25. Starting at the lower model level the Richardson number in both methods is determined at successively greater heights by use of linear interpolation between adjacent model levels. The boundary-layer top is assigned to the height where the Richardson number exceeds a given value. Both expressions for the Richardson number are proportional to z(θ(z) − θ(s)). In the ideal case where the virtual potential temperature is constant in the boundary layer and increases at a certain rate above it, this means that for increasing z, a correspondingly smaller temperature change is needed in order to reach the prescribed
36
SVEN-ERIK GRYNING AND EKATERINA BATCHVAROVA
Richardson-number value. This makes the determination of the boundary-layer height sensitive to even small changes between successive temperature profiles, and may partly explain the large variability that is often found in time series of the boundary-layer height determined from numerical weather-prediction models by use of the Richardson-number method. The expressions treat differently the wind-velocity influence. In Equation (1) the wind speed is taken at the given height, whereas Equation (2) applies the difference between the lowest model level and the actual height, and the surface layer is accounted for through an additional friction-velocity term. This term can be large compared to the wind-profile contribution. Then the boundary-layer height is determined mainly from the temperature profile and the friction velocity. Over water, owing to the small roughness length, the wind speed is typically high with small friction velocity. Hence over water the Richardson number suggested by Sørensen (1998) would tend to predict a higher boundary layer as compared to the Richardson number suggested by Vogelezang and Holtslag (1996). 3.2. S IMPLE APPLIED MODEL The height of the boundary layer was simulated with a high-resolution applied slab-type model. Gryning and Batchvarova (1996) base this model on a zero-order scheme for the development of the (internal) boundary layer during near-neutral and unstable atmospheric conditions. The energy balance equation within the boundary layer and the potential temperature jump at the top of it are parameterised; initially the boundary-layer growth is proportional to friction velocity. As the layer grows the production of mechanical turbulent kinetic energy controls the process, and the production of convective turbulent kinetic energy becomes increasingly important with deepening of the layer and increasing turbulent sensible surface heat flux. The equation for the height h of the internal boundary layer (Gryning and Batchvarova, 1996) is Cu2∗ T h2 + (1 + 2A)h − 2BκL γ g[(1 + A)h − BκL] ∂h ∂h (w θ )s ∂h +u +v − ws = , (3) × ∂t ∂x ∂y γ where u and v are the horizontal components of the mean wind speed in the internal boundary layer in the x and y directions, t is time, L is the Obukhov length, κ is the von Karman constant, and A, B and C empirical dimensionless constants from a parameterisation of the turbulent energy production equation. Here we use A = 0.2, B = 5 and C = 8. The input to the model consists of wind speed and direction (or the wind vector (u, v)), the friction velocity u∗ , the kinematic heat flux (w θ )s in time and space for the modelling domain, and the potential temperature gradient γ and the mean vertical air motion above the boundary layer ws as a function of height
MARINE BOUNDARY LAYER AND TURBULENT FLUXES OVER THE BALTIC SEA
37
and time. The model includes coastline curvature and spatially varying winds and was solved numerically for the height of the internal boundary layer.
4. Model Simulations Both models were applied for the period 26 October to 2 November 1998. 4.1. HIRLAM MODEL The hourly near-surface wind field and fluxes of latent and sensible heat from the HIRLAM model represent mean values over the 22.5 km × 22.5 km grid cells and do not contain smaller scale information. The fluxes of latent and sensible heat, and for use in the simple applied model the u and v wind field components, were interpolated to a 2 km × 2 km grid. Inverse distance squared interpolation between the model predictions at all the HIRLAM grid points was used. 4.1.1. Wind Field The interpolated wind speed and direction predicted by the HIRLAM model at Christiansø during the intensive period are shown in Figure 5, together with the measured values. The HIRLAM wind speed is representative for a height of 10 m, and was derived, together with the friction velocity, from the lowest model level by use of a parameterised wind profile. At the start of the period the wind speed is very high, followed by moderate values at the end. The HIRLAM model successfully reflects the changes in wind direction and speed. The agreement between modelled and measured wind speed and direction is indeed very good, with deviations between model and measurements typically 20◦ and 2 m s−1 . It should be noted that for the sector 190◦ to 270◦ the air flow has passed over Bornholm some 20 km away. In the northerly sector the water fetch to Christiansø is of the order of 100 km or more. 4.1.2. Friction Velocity and Fluxes of Sensible and Latent Heat Figure 6 shows the predictions from the HIRLAM model and measurements of sensible and latent heat fluxes for the period 26 October to 3 November, and Figure 7 presents the air temperature and the measured and predicted friction velocity. At high wind velocities the fluxes of sensible and latent heat are typically between zero and 100 W m−2 , except for two cases with very high sensible heat flux on 27 and 28 October. The fluxes show variations on both small and large time scales, but the daily variation that is characteristic for fluxes over land surfaces is not present. For the friction velocity and the sensible heat flux, the agreement between HIRLAM and measurements is generally good, except for a few cases with very high values of both the measured sensible heat flux and friction velocity. The analysis of data and model results revealed that these high values correspond to
38
SVEN-ERIK GRYNING AND EKATERINA BATCHVAROVA
Figure 5. Wind speed and direction during the period 26 October to 3 November 1998. The time is shown in hours, starting on midnight 25/26 October and ending on midnight 3/4 November 1998 (hour 216). The full line represents measurements at Christiansø, the dashed line the interpolated predictions by the HIRLAM model. The upper panels show the time evolution, the lower panels are scatter plots.
short-duration occurrences of high wind speeds at Christiansø (around 40 and 60 hours on Figure 5). The HIRLAM model poorly resolved these periods, also for wind speed. For the latent heat flux, the measurements are clearly lower than the predictions by HIRLAM. 4.1.3. Boundary-Layer Height Two Richardson-number methods to extract the boundary-layer height from the output of numerical weather prediction (NWP) models were applied to the hourly output from the HIRLAM model. The application is illustrated in Figure 8 for several cases. The virtual potential temperature (middle panels) decreases in the layer near the ground and then increases slowly as a function of height. Near the top of the boundary layer the increase becomes larger. The behaviour of the Richardsonnumber profile is alike; near the ground the Richardson number from Equation (1)
MARINE BOUNDARY LAYER AND TURBULENT FLUXES OVER THE BALTIC SEA
39
Figure 6. Same as in Figure 5, but for the sensible and latent heat fluxes.
Figure 7. Left panel shows measurements of air temperature from the sonic anemometer and the two bullets show the only available bucket measurements of the water temperature. In the right panel the full line shows measured friction velocity and the dashed line the predictions by the HIRLAM model. Time indications as in Figure 5.
40
SVEN-ERIK GRYNING AND EKATERINA BATCHVAROVA
is slightly negative and then increases significantly in response to the large increase of temperature. The Richardson number from Equation (2) shows a similar behaviour but with larger variability, being more negative near the ground and having a much larger increase in response to the temperature increase near the top of the boundary layer. It can be seen that the vertical grid resolution in the HIRLAM model is too coarse to justify a subjective determination of the boundary-layer height, following the procedure for radiosoundings. The analysis covers four gridpoints close to Christiansø, but the predicted boundary-layer heights were so similar that only results from the grid point closest to Christiansø will be reported, see Figure 9. The result from the analysis using the Richardson number suggested by Sørensen (1998) is shown on the left panel. It can be seen that the predicted boundary-layer height is clearly too high during the first part of the experimental campaign where the wind is southwesterly. At 160 hours when the wind turns north, such that Bornholm no longer affects the air mass over Christiansø, agreement between measurements and predicted boundary-layer heights improves considerably. The right panel shows the results when using the Richardson number suggested by Vogelezang and Holtslag (1996). It can be seen that the predicted boundary-layer height generally is substantially lower than on the left panel, but still overpredicts the boundary-layer height for the first part of the simulation where the wind passes over Bornholm before reaching Christiansø. There is nice agreement during the last part when the wind is northerly and the effect of Bornholm is absent. 4.2. S IMPLE APPLIED MODEL The evolution of the boundary layer over Christiansø for the period 26 October to 2 November 1998 was simulated with the simple applied model. The model domain is shown in Figure 1. It extends 800 km both in the west-east and the south-north direction. A grid resolution of 2 km and a time step of 15 s were used. The simulations were performed in two ways. One was based on the measured wind field at Christiansø, which was assumed to be representative for the whole model domain. The other simulation was based on the wind field from the HIRLAM model. Interpolated u and v wind field components at each grid point in the 2-km grid are derived for the whole model domain by inverse square interpolation between the wind predictions at the HIRLAM grid points. For both simulations the sensible heat flux and the friction velocity measured at Christiansø were used for the entire modelling domain. These parameters are expected to vary within the Baltic Proper. However the wind during the experimental campaign was between southwest and north, corresponding to a water fetch to Christiansø of the order of 100 km. Under these conditions we assume the observations at Christiansø to be representative for the upwind sea-surface area. Subsidence was neglected in both cases.
MARINE BOUNDARY LAYER AND TURBULENT FLUXES OVER THE BALTIC SEA
41
Figure 8. Examples of HIRLAM model output and the corresponding Richardson numbers. Left panels show the wind-speed components, () u-component (easterly) and () v-component (northerly). The central panels are the virtual potential temperature, θv , and the right panels profiles of the Richardson numbers, () Equation (1) by Sørensen (1998) and () Equation (2) by Vogelezang and Holtslag (1996).
42
SVEN-ERIK GRYNING AND EKATERINA BATCHVAROVA
Figure 9. Boundary-layer height over Christiansø during the extensive observation period, estimated from profiles of wind and temperature from the HIRLAM model are shown by thin lines. The results represent the HIRLAM grid point closest to Christiansø. The left panel shows the results using the Richardson number suggested by Sørensen (1998), the right panel when using the Richardson number in Vogelezang and Holtslag (1996). Bullets show measurements. The thick lines illustrate a running mean over 9 points. Time indications as in Figure 5.
Figure 10 shows the evolution of the simulated boundary-layer height over Christiansø, when taking into account the effect of Bornholm. The left panel shows the results when the measured wind field at Christiansø is used in the simulation, the right panel when the interpolated HIRLAM wind field is used. It can be seen that the simulated height of the boundary layer over Christiansø is only marginally dependent on the choice of wind field. Large differences between the simulations can be observed, however, for wind directions close to 270◦ when one of the wind fields reflects that the air flow passes over Bornholm and the other that the air does not pass Bornholm but has a very long water fetch. The bullets show the observations of the boundary-layer height from the radiosoundings; the overall agreement is fairly good. Figure 11 similarly shows the evolution of the boundary-layer height but without accounting for the effect of Bornholm; Bornholm is replaced by water. As before, the difference between simulations based on the measured wind field at Christiansø and those based on the wind field from the HIRLAM model is small. It is, however, very clearly seen that for the period up to about 160 h after the start of the simulation, the simulated boundary layer is markedly higher than the measured one. After 160 h of simulation the agreement becomes fair. Inspecting the wind direction reveals that up to 160 h, the wind is within the sector that includes Bornholm, and around 160 h the wind turns towards north and the air that reaches Christiansø has not passed over Bornholm but originates from the Swedish coast.
MARINE BOUNDARY LAYER AND TURBULENT FLUXES OVER THE BALTIC SEA
43
Figure 10. Simulation with the simple applied model of the evolution of the boundary-layer height over Christiansø during the extensive observation period. In the simulations the effects of Bornholm is accounted for. Bullets show measurements. In the left panel, the wind measurements at Christiansø are used in the simulation, in the right panel, the interpolated HIRLAM wind field for the whole area is used. Time indications as in Figure 5.
5. Results and Discussion During the experiment the water was generally warmer than the air (Figure 7), which is a very typical feature for the Baltic Sea during the late summer, autumn and early winter. This results in a generally positive sensible heat flux to the atmosphere, as can be observed for the experimental period in Figure 6, and the generation of a convectively driven boundary layer over the water. The experimental period covers wind speed in the range from calm to 19 m s−1 . The period from 26 October until midday 1 November 1998 is characterised by winds about 12 m s−1 from southwest to west. In this sector Christiansø is downwind of Bornholm with a water fetch of about 20 km. Following a wind direction shift on 1 November 1998 to northwest and north, the wind decreased to about 4 m s−1 . Then Christiansø is not downwind of Bornholm and the over water fetch from the Swedish coast is about 100 km. The model simulations suggest that during the first period the island of Bornholm controls the boundary layer over Christiansø. The high-resolution simple applied model successfully predicts the boundary-layer height over Christiansø, but it vastly overestimates it when the effect of Bornholm is omitted from the simulation. The boundary-layer height that was estimated from the HIRLAM data by use of the Richardson-number methods is also higher than the measured one, which suggests that the HIRLAM model did not resolve the mesoscale features that control the boundary-layer height over Christiansø. During the last period of the experiment the wind was northerly. The air that reached Christiansø did not pass Bornholm on its way, but the water fetch to the
44
SVEN-ERIK GRYNING AND EKATERINA BATCHVAROVA
Figure 11. Simulation with the simple applied model of the evolution of the boundary-layer height over Christiansø during the extensive observation period. In the simulations Bornholm is replaced by water. Bullets show measurements. In the left panel, the wind measurements at Christiansø are used in the simulation, in the right panel, the interpolated HIRLAM wind field for the whole area is used. Time indications as in Figure 5.
nearest coast was about 100 km. For this case good agreement between measured and simulated boundary-layer heights was found for all the model simulations. It should be noted that the predictions of the boundary-layer height based on the HIRLAM data do not only reflect differences in the two Richardson-number schemes but also depend on the degree of sophistication of the parameterisations of the boundary-layer parameters used in the HIRLAM model, and on the vertical resolution of the boundary layer within the model. For the sensible heat flux the agreement between measurements and predictions by the HIRLAM model is generally good. However some high wind speed situations were not resolved satisfactorily, which results in an under prediction of the larger sensible heat fluxes. For the latent heat the measurements are clearly lower than the predictions by the HIRLAM model. These findings are in agreement with Rutgersson et al. (2001), who show a comparison between results from the HIRLAM model, the ocean PROBE-Baltic model (Omstedt and Nyberg, 1996), and measurements from Christiansø, covering the period May to December 1998. Rutgersson et al. (2001) found that the HIRLAM model overestimated the latent heat flux at Christiansø as compared to the measurements. The effect was partly ascribed to an excessively dry surface layer predicted by the HIRLAM model, which indicates too much land influence and likely is due to the large scales in the model and the SHMI (1 × 1)◦ database. The HIRLAM model also overestimated the sensible heat flux but to a much lesser degree. The relatively large deviations between model and measurements show that heat and water budgets from models should be interpreted with caution in the vicinity of coastlines.
MARINE BOUNDARY LAYER AND TURBULENT FLUXES OVER THE BALTIC SEA
45
6. Conclusions During the intensive experimental period (26 October to 3 November 1998), the meteorological conditions at Christiansø were characterised by an upward directed heat flux over the sea, and strong westerly winds, which on November 1st decreased and turned towards north. Under these meteorological conditions in was found, that: 1. The marine boundary-layer height was typically 500 m over Christiansø. 2. The marine boundary-layer height over Christiansø varies on small and long time scales, but does not show the daily variation that is typical over land. 3. Bornholm influences the boundary-layer height over Christiansø for winds in the sector south to west where Christiansø lies downwind of Bornholm. The water fetch between Bornholm and Christiansø is about 20 km. 4. A simple applied, high resolution slab-type model of the boundary-layer height reproduced the characteristic behaviour of the boundary layer over Christiansø, both when Christiansø lies downwind of Bornholm with a water fetch of 20 km, and in northerly winds when Christiansø is downwind of the Swedish coast with a water fetch of 100 km or more. 5. The boundary-layer height was estimated from HIRLAM model output. Two methods were applied, both based on a Richardson-number approach. Both methods worked well for northerly winds, but both failed when Christiansø was downwind of Bornholm. 6. The grid resolution in the HIRLAM model is of the same size as the distance between Christiansø and Bornholm and the size of Bornholm itself. It is too coarse to reflect the mesoscale features that control the boundary-layer height over Christiansø when Christiansø is downwind of Bornholm. It seems to be adequate for northerly winds when Christiansø is downwind of the Swedish coast with a water fetch of 100 km or more. 7. The HIRLAM model as compared to measurements at Christiansø overestimates the latent heat flux. The sensible heat flux and friction velocity are fairly well predicted.
Acknowledgements We are thankful to the Swedish Meteorological and Hydrological Institute for generously providing the data from the HIRLAM simulations, and it is a special pleasure to acknowledge fruitful and constructive discussions with Anna Rutgersson. The sea surface temperature data at Christiansø are from the Danish Meteorological Institute. The advice and help of G. Vestergård Pedersen, P. Holmgaard Christensen, and J. Kidholm Christensen, Administrators of Ertholmene during various parts of the project, are very much appreciated. Also we wish to express our gratitude to the
46
SVEN-ERIK GRYNING AND EKATERINA BATCHVAROVA
people of the islands for their supportive interest in our work. We thank S. Bockelund for her informative article on our research in the local newspaper (Bornholms Tidende) and T. Andersen, teacher at the local school, for inviting us to talk about meteorological soundings, followed by practical exercises. The study is a part of the European Union supported PEP-in-BALTEX project, contract number ENVC4-CT97-0484. Ekaterina Batchvarova carried out part of the work while being a visiting scientist at the Joint Research Centre, Ispra.
References Batchvarova, E. and Gryning, S.-E.: 1998, ‘Wind Climatology, Atmospheric Turbulence and Internal Boundary-Layer Development in Athens during the MEDCAPHOT-Trace Experiment’, Atmos. Environ. 32, 2055–2069. Batchvarova, E., Cai, X., Gryning, S.-E., and Steyn, D.: 1999, ‘Modelling Internal Boundary-Layer Development in a Region with a Complex Coastline’, Boundary-Layer Meteorol. 90, 1–20. Brooks, I. M., Goroch, A. K., and Rogers, D. P.: 1999, ‘Observations of Strong Surface Radar Ducts over the Persian Gulf’, J. Appl. Meteorol. 38, 1293–1310. Dorman, C. E.: 1994, ‘Guadalupe Island Cloud Trail’, Mon. Wea. Rev. 122, 235–242. Garratt, J. R.: 1987, ‘The Stably Stratified Internal Boundary Layer for Steady and Diurnally Varying Offshore Flow’, Boundary-Layer Meteorol. 38, 369–394. Garratt, J. R.: 1990, ‘The Internal Boundary Layer – A Review’, Boundary-Layer Meteorol. 50, 171–203. Gryning, S.-E.: 1985, ‘The Øresund Experiment – A Nordic Mesoscale Dispersion Experiment over a Land-Water-Land Area’, Bull. Amer. Meteorol. Soc. 66, 1403–1407. Gryning, S.-E. and Batchvarova, E.: 1996, ‘A Model for the Height of the Internal Boundary Layer over an Area with a Irregular Coastline’, Boundary-Layer Meteorol. 78, 405–413. Gryning, S.-E. and Joffre, S.: 1987, ‘Wind Structure over the Øresund Strait’, in The Øresund Experiment, Proceedings from Workshop II, Uppsala, October 13–14 1987. Available from the Library, Risø National Laboratory, DK-4000 Roskilde, Denmark, pp. 11–19. Hsu, S. A.: 1983, ‘On the Growth of a Thermally Modified Boundary Layer by Advection of Warm Air over a Cooler Sea’, J. Geophys. Res. 88, 771–774. Hsu, S.A.: 1984, ‘Effect of Cold-Air Advection on the Internal Boundary-Layer Development over Warm Oceanic Currents’, Dyn. Atmos. Oceans 8, 307–319. Hsu, S. A.: 1988, Coastal Meteorology, Academic Press, San Diego, 260 pp. Källén, E.: 1996, Hirlam Documentation Manual, System 2.5, SMHI, SE-601-76 Norrköping, Sweden, 180 pp. Källstrand, B. and Smedman, A.-S.: 1997, ‘A Case Study of the Near-Neutral Coastal Internal Boundary-Layer Growth: Aircraft Measurements Compared with Different Model Estimates’, Boundary-Layer Meteorol. 85, 1–33. Melas, D.: 1998, ‘The Depth of the Stably Stratified Internal Boundary Layer over the Sea’, Geophys. Res. Lett. 25, 2261–2264. Melas, D. and Kambezidis, H. D.: 1992, ‘The Depth of the Internal Boundary Layer over an Urban Area under Sea-Breeze Conditions’, Boundary-Layer Meteorol. 61, 247–274. Moszkowicz, S., Ciach, G. J., and Krajewsky, W. F.: 1994, ‘Statistical Detection of Anomalous Propagation in Radar Reflectivity Patterns’, J. Atmos. Oceanic Tech. 11, 1026–1034. Mulhearn, P. J.: 1981, ‘On the Formation of a Stably Stratified Internal Boundary Layer by Advection of Warm Air over a Cooled Surface’, Boundary-Layer Meteorol. 21, 247–254.
MARINE BOUNDARY LAYER AND TURBULENT FLUXES OVER THE BALTIC SEA
47
Omstedt, A. and Nyberg, L.: 1996, ‘Response of Baltic Sea Ice to Seasonal Interannual Forcing and Climate Change’, Tellus 48A, 644–662. Rutgersson, A., Smedman A.-S., and Omstedt, A.: 2001, ‘Measured and Simulated Latent and Sensible Heat Fluxes at Two Marine Sites in the Baltic Sea’, Boundary-Layer Meteorol. 99, 53–84. Sørensen, J. H.: 1998, ‘Sensitivity of the DERMA Long-Range Gaussian Dispersion Model to Meteorological Input and Diffusion Parameters’, Atmos. Environ. 24, 4195–4206. Stull, R. B.: 1983, ‘Integral Scales for the Nocturnal Boundary Layer, Part 1: Empirical Depth Relationships’, J. Clim. Appl. Meteorol. 22, 673–686. Vihma, T. and Brümmer, B.: 2002, ‘Observations and Modelling of the On-Ice and Off-Ice Air Flow over the Northern Baltic Sea’, Boundary-Layer Meteorol. 103, 1–27. Vogelezang, D. H. P. and Holtslag, A. A. M.: 1996, ‘Evaluation and Model Impacts of Alternative Boundary-Layer Height Formulations’, Boundary-Layer Meteorol. 81, 245–269.