3.30

Offshore wind modelling and forecast J. Beran∗,L. Calveri, B. Lange, Lueder von Bremen FORWIND, University Oldenburg, Germany June 6, 2005

Abstract

The conditions offshore differ significantly to that known over the land. Thermal inertia of water one order of magnitude higher than of the soil, the short wave radiations propagation in depth of a water column and a buoyancy driven turbulence result in very slow heating of first couple of meters of water column with no diurnal cycle. However, the sea temperature changes throughout the year with a slight phase shift to the annual air temperature. The flexible sea surface interacts with the overflowing wind Csanady (2001) producing a wavy sea, which modifies the lower boundary parameters and the flow above. When an well developed planetary boundary layer is advected over the colder sea, the PBL accommodates to new conditions and an internal boundary layer (hereafter IBL) develops. Negative buoyancy flux neglects the shear driven turbulence flux and remaining turbulence decays due to omnipresent molecular diffusion. In case of sharp temperature gradients, the IBL decouples from the decaying relict PBL and without turbulence driven friction, the balance of geostrophic forces is lost. The wind starts to be supergeostrophic and an inertial oscillation is initiated Kallstrand (2000). The inertial oscilation was derived by Blackadar as

Offshore wind power projects are critically reliant on accurate wind resources assessment and large offshore windfarms require timely weather forecast. Most often, however, the offshore measurement is scarce and conventional large scale weather modelling with typical resolution 0.5◦ x 0.5◦ is rather limited in application. The lack of measurement and the need for a high resolution weather forecast can be tackled by using a Limited Area Model, provided that long term validation of the model is performed. We use the MM5 to model and forecast offshore winds, while conducting series of validation.

1. Introduction The limited area model (hereafter LAM) is designed to forecast only mesoscale features, which are superposed to the synoptic weather pattern, thus LAM can be used to refine and improve the forecast in regions, where mesoscale is significant or higher resolution is needed. Such a region can be found over the sea, where measurements are not widely available and close to the coastline, where the sharp transition in the underlying boundary has an influence on the flow. Additionally, there are some particularities of the offshore weather originating from different physical and geometric properties of lower boundary. Whilst the sea breeze is rather weaker effect in higher latitude Kallstrand (2000), the most essential is the stable boundary layer decoupling and occurrence of low-level jet Smedman (1997).

∂ (u − ug ) = f (v − vg ) ∂t ∂(v − vg ) = −f (u − ug ) ∂t which, after rearanging in complex form

V = (ug − u) + i(vg − v) ∂V = −if V , ∂t has the complex solution

V = V0 e−if t

∗ [email protected]

1

WRF/MM5 Users' Workshop - June 2005

.

(1)

(2)

(3)

(4)

(5)

By assuming two Ekman profiles, varying each from other only in turbulent diffusion coefficient Kturb the hodograph in figure 1 can explain the initial perturbation V0 , when the frictional force reduces suddenly. Theoretical Ekman spiral 4

Kturb = 5 m22/s Kturb = 1 m /s 200m

3

V-wind [m/s]

50m

200m

2

50m 1

4

2

0

10

8

6 U-wind [m/s]

Theoretical Ekman spiral difference 2

Difference 50m

V-wind [m/s]

1

0

100m

800m 200m

-1

-2

400m -3

-1

0

2 1 U-wind [m/s]

3

4

Figure 1: Ekman spiral and differenced Ekman spirals.

There are two problems arising from such a simplified formulation. The turbulent diffusion coefficient Kturb being a property of the turbulent flow is not known apriori and secondly the assumption of sudden step in turbulent diffusion is not in agreement with the existence and growth of IBL Kallstrand (2000). For this reason, a complex model has to be used. Such a model should resolve in a great detail the horizontal temperature gradient and should be able to interact with a global model to simulate a real weather.

WRF/MM5 Users' Workshop - June 2005

For the reason of user support, open source policy and versatility, we have chosen the fifth generation mesoscale model MM5 to simulate and forecast offshore weather features, which are of subgrid scale in a global weather modelling systems. The MM5 uses the full compressible set of continuity equation, Navier-Stokes equations and energy equation. The numerical solution is computed in the rectangular terrain following structured staggered 1 grid by finite difference schemes, while using Leap-frog time marching. We retrieve and use FNL 1◦ x 1◦ analysis as the initial and lateral boundary conditions from Data Mass Storage at NCEP Shea (1995). We have also been using upper air and surface observation to compare our results, though there are very scarce measurement offshore, mostly from light ships along traffic coridors, oil drilling rigs and a occasional measurement campaign. An unique opportunity to validate our results has been provided recently on 100m high offshore research platform FINO-1 in the North Sea.

0

-1

2. Mesoscale model MM5 and configuration

Warner (1997) suggests that main LAM errors are associated with the LBC imposed on laterals of LAM domain. However, the LBC coming for FNL are too coarse to be directly imposed at the laterals of the very resolved domain. According to Warner, the LBC are downscaled using several buffer domains between global model input and area of interest. We have addressed this issue and we have run MM5 in several configuration over a period of three days. Two periods were selected on 27th-30th October 2003 and 14th-17th March, with unstable and stable stratification respectively. Secondly, the vertical structure of the grid was varied. In fact, it is very difficult to select the right vertical structure for stable conditions as the height of low level jet maximum is a part of solution. This means that the vertical grid structure should be very dense in the lowest part of domain below the initial PBL height. Three different stretching of the vertical grid used in the work, are shown in figure 2. 1 Arakawa-Lamb

B-staggered grid

Profiles mean

2

Vertical coordinate stretching

10

500

300 log(height)[m]

Height [m]

400

200

prof FINO prof 3Dom 2Way prof 3Dom 1Way prof 2Dom 2Way prof 2Dom 1Way prof 2Dom 1Way 60 liv prof 2Dom 1Way 80 liv prof 1Dom

100

0

40 levels

60 levels Setup

80 levels

Figure 2: Vertical grid stretching in PBL.

3. Results of MM5 run To answer the LBC error propagation, the resolution of finest nested domain has been combined (3 and 10 km) with both ways of nesting, whilst all other parameters has been kept constant. The results show that the best trade off is reached when using 10 km resolution one-way nesting with one mother domain at horizontal resolution 30 km. This is believed to be due to rather homogeneous off-shore conditions with no orographic forcing at all. The increased number of vertical levels has shown to improve the solution when compared to the measurement, though there is still visible dissagreement with the measured wind profile at 100m. It is particularly important to increase the amount of vertical levels above the log layer, in other words in the height where LLJ could occur. The resulting vertical profile of the stable spring period model runs with variable vertical and horizontal resolution are shown in figure 3 and compared to measurement at FINO1 research platform. In contrary to the vast measurement network on land, there are very scarce measurements offshore. It makes validation of the model very troublesome, even though the situation was eliviated by a recent economic interest leading to construction of several measurement masts in the sea. Although some masts are extending up to 100 meters above the MSL, there is still only indirect evidence of the LLJ and similar phenomena in the marine PBL. The

WRF/MM5 Users' Workshop - June 2005

11

12

13

15 14 wind speed [m/s]

16

17

18

Figure 3: Stable strat. wind profiles 14-17 March 04.

only way up to this date available to indicate offshore LLJ is the upper-air measurement at oil drilling rigs in North Sea, apart of numerical results. Favourable condition for an occurance of LLJ has happened on 15-17th April 2005, when the continental unstable PBL has been blown over much colder North Sea with a significant temperature gradient ≈ 5K/100km accros the coastline. The model geostrophic wind and 150m wind fields are shown in figure 4, where central area offshore shows significant increase in speed due to IBL decoupling. The area matches exactly the finest computational domain, and the influence of the coarser domain could be seen on boundaries, where wind speeds are depreciated from laterals. The example of the LLJ oocurence predicted by MM5 is shown here for ilustration and it is believed that the further research will show, how important it is for all offshore operations. The figure 5 shows the vertical profile of the wind in lower part of atmosphere at FINO-1 research platform. However, the offshore area affected by LLJ seems to be much larger as the significant low level overspeeding is still noticable at 250 km distant oil rig Ekofisk downstream from the FINO-1 shown in figure 6. In fact, the oil rig is literally in the middle of the North Sea, which could also mean, that LLJ can cover large areas and could

Figure 4: Geostrophic wind (arrows) and 150m wind (contours) fields 00UTC 18Apr05.

be initiated under favourable condition on all North Sea coasts.

4. Conclusion and recommendation

futher

work

The apparent homogenity of sea surface provides great advantage for LAM simulation, where the horizontal domain resolution can be significantly coarser than resolution used in complex terrain. Additionally, one way nesting is considered satisfactory in offshore conditions without any orographic forcing. By increasing the vertical grid structure, the model results were improved slightly, though the vertical velocity profile is still far from being perfect when compared to measurements. However, the time constraint allowed us to evaluate only short periods and also the model accuracy in coastal areas has not yet been investigated. The results with a good correlation to offshore measurements seems to be promising signal for a future work. The existence of IBL decoupling and occurence of LLJ regardless on the distance and wind direction could be more regular and far from seldom. There is hope that future work will improve the understanding of this phenomenon.

WRF/MM5 Users' Workshop - June 2005

Figure 5: MM5 (above) and GFS (below) vertical profile of the wind in lower atmosphere.

Warner, P. R. T. R., T. T., 1997: A tutorial on lateral boundary condition as a basic potentially serious limitation to regional numerical weather prediction. Bull. of the Amer. Mer. Soc., 78, 2599–2617.

Figure 6: Wind profile at Ekofisk research rig.

Acknowledgement The authors hereby acknowledge support from the EC ”Wind Energy Assessment Studies and Wind Engineering” (WINDENG) Training Network (contract n. HPRN-CT-2002-00215).

References Csanady, G., 2001: Air-Sea Interaction / Law and Mechanisms. Kallstrand, B. H. H. J. S. A., B., 2000: Mesoscale wind field modification over teh baltic sea. Boundary Layer Meteorology, 95(2), 161–188. Shea, W. S. S. I. H. J., D.J., 1995: An Introduction to Atmospheric and Oceanographic Datasets. National Center for Atmospheric Research (NCAR), Bolder, USA. Smedman, B. H. G. B., A.-S., 1997: Evolution of stable boundary layers over a clod sea. Journal of Geographical Research, 102, 1091–1099.

WRF/MM5 Users' Workshop - June 2005