NMOC Quarterly Summary April-June 2002

Marine surface Winds from LAPS

Analyses & Numerical Prediction | About Products | Map/Image/Chart Archives

Introduction

In 1999, there was a major upgrade in the configuration of the Bureau's regional forecasting model, LAPS. The changes included an increase in resolution from 0.75° and 19 levels to 0.375° and 29 levels with more levels in the boundary layer, and changes to parameterisations, most notably, the replacement of the boundary-layer scheme by that used by the ECMWF. Since the introduction of the new model, there has been observed a negative bias in the marine winds from the LAPS system, evident in the performance of the wave models forced by these winds and from feedback from operational forecasters comparing modelled winds with observed winds from coastal and ship reports. In order to clarify the issue, several meetings were convened to look at various aspects of the model configuration and performance.^*

Wave Model Response

As presented in these bulletins, the verification of significant wave height (SWH) forecasts from the wave models against observations from waverider buoys off the Australian coast generally shows a negative bias in the wave forecast from the regional model (forced by LAPS winds). This is a change in performance from 1999 when there was negligible bias and even overforecasting of waves in some significant events.

For a fully developed sea-state, the SWH is proportional to the square of the 10 metre wind speed. The wave model is therefore a sensitive indicator of potential errors in the surface wind fields used to force the model. As the wave model was unchanged when the LAPS changes were introduced, the change in behaviour must be attributed to the wind forcing derived from new configuration of the model. What is not clear is whether this is due to the change in the boundary layer formulation or due to model biases in the intensity of weather systems (which may be caused by a number of reasons such as model resolution, physical parameterisations and initial conditions).

Wind Verification

The main problem with verifying marine surface winds is identifying suitable observations that are representative of the marine environment. Anemometers at most coastal stations are sited too far inland, ship-based observations are unreliable and buoys often have their anemometer close to the surface so that it is partly shielded by waves. Although close to the coast, the Sydney Ocean Reference Station and Oil Platforms at North Rankin and Kingfish B have been identified as suitable. In the broader domain, there are several platforms, including the Tropical Atmosphere Ocean (TAO) array, providing observations of marine winds but these are not particularly useful for verification of LAPS.

Verification of surface winds against the platforms in the Australian region presented in this bulletin consistently shows a negative bias in the wind speed. These results may, however, be slightly contaminated by the fact that land points (where the wind is known to be too light) are included in the interpolation.

Jeff Kepert has compared LAPS forecast marine winds with scatterometer data from QuikScat. Scatterometer data are ideal for such a comparison, as it is available over the whole of the marine part of the model domain, and apart from some known shortcomings, it is highly accurate. It is not currently used in the operational LAPS suite, so is an independent test. Results show that there is a slight (1m/s) negative bias of model wind speed relative to the scatterometer winds. About half of this is due to poor quality scatterometer data due to rain contamination or being near the centre or edge of the swathe. Fitting a regression curve to the data shows that model winds are roughly 10% lighter than the (reliable) scatterometer data, with the bias being more marked in the meridional component than in the zonal.

Comparisons by Diana Greenslade of LAPS surface winds with co-located GASP winds show that for low wind speeds (< 5m/s) there does not appear to be any significant bias between LAPS and GASP. However, for higher wind speed there is a tendency for LAPS to underestimate wind speed compared to GASP.

Synoptic Assesment

In late July of 2001 an intense mid-latitude cyclone developed off the NSW coast and produced strong southerly gales in the coastal waters. This storm coincided with a yacht race from Sydney to Southport and produced a significant amount of public interest. Operational forecasters in Sydney reported that although the LAPS model guidance was very good, the model winds at the sigma=0.9875 level (about 100 metres) compared better with observations than the 10 metre winds.

Jeff Kepert compared the model forecasts for this period with QuikScat data. The results show that the period of greatest negative bias of the model coincided with a particularly poor MSLP forecast, with the near coastal southerly gradient substantially under predicted. In this case it is mainly the pressure gradient that contributes to the low surface wind bias rather than the boundary layer reduction to 10 metres.

Model Changes

In the boundary layer formulation the surface (10 metre) winds are derived from a roughness length for momentum based on the Charnock parameter. This was assumed to be constant and equal to 0.018 based on observations which were probably dominated by limited fetch and shallow water. Observations during TOGA-COARE, which are thought to be more representative of the open ocean, suggest a value of 0.011. This has been changed in the operational version of the model but seems to have had minimal impact. Dale Hess is now suggesting a variable Charnock parameter dependent on wind speed.

Kamal Puri has suggested a reduction in horizontal diffusion that would increase the depth of low pressure systems and strengthen surface winds. In trials to date, S1 skill scores of the resulting forecasts show an improvement (at all forecast times) over the current operational model, particularly as a consequence of a more realistic depiction of the depth of cut off lows.

Summary

It has been agreed that there is small, but noticeable negative bias in the marine surface winds from LAPS. It still not clear if this is due to systematic errors in the boundary-layer formulation or to errors in the model forecasts. As the major errors do seem to be at higher wind speeds it is likely that model errors in depths of systems is a major contributing factor.
Both issues are being addressed with changes being made to the boundary layer parameterisation and to the horizontal diffusion to try to produce more intense systems in the forecasts. Future work will involve more rigorous verification of the surface winds against in situ data that is available in all model domains, and possible ongoing verifications of the operational forecasts against QuikScat data.

Reference

Desjardins, S. et al., 2000: Examination of the impact of a coupled atmospheric and ocean wave system. Part I: Atmospheric aspects. J. Phys. Oceanog. 30. 385-401