Search Results

You are looking at 1 - 4 of 4 items for :

  • Author or Editor: Diana Greenslade x
  • Weather and Forecasting x
  • Refine by Access: All Content x
Clear All Modify Search
Frank Woodcock
and
Diana J. M. Greenslade

Abstract

The operational consensus forecast (OCF) scheme uses past performance to bias correct and combine numerical forecasts to produce an improved forecast at locations where recent observations are available. Here, OCF uses past observations and forecasts of significant wave height from five numerical wave models available in real time at the Australian Bureau of Meteorology. In addition to OCF, different adaptive weighting and forecast combination strategies are investigated. At deep-water sites (ocean depth > 25 m), all of the interpolated raw model forecasts outperformed 24-h persistence and, after bias correction, one model was clearly best. Significant improvements over raw model significant wave height forecasts were achieved by bias correction, linear-regression methods, and combination strategies. The best forecasts were obtained from a “composite of composites” in which models with highly correlated errors were combined before being included in the performance-weighted bias-corrected forecast. This technique slightly outperformed the linear-regression-corrected best model. At shallow-water sites (ocean depth < 25 m), all raw models perform poorly relative to the 24-h persistence. The composited, corrected forecasts significantly improved on raw model significant wave height forecasts but only slightly outperformed the 24-h persistence. The raw models generated unrealistically large biases that tended to be amplified with larger observed values of significant wave height.

Full access
Eric W. Schulz
,
Jeffrey D. Kepert
, and
Diana J. M. Greenslade

Abstract

A method for routinely verifying numerical weather prediction surface marine winds with satellite scatterometer winds is introduced. The marine surface winds from the Australian Bureau of Meteorology’s operational global and regional numerical weather prediction systems are evaluated. The model marine surface layer is described. Marine surface winds from the global and limited-area models are compared with observations, both in situ (anemometer) and remote (scatterometer). A 2-yr verification shows that wind speeds from the regional model are typically underestimated by approximately 5%, with a greater bias in the meridional direction than the zonal direction. The global model also underestimates the surface winds by around 5%–10%. A case study of a significant marine storm shows that where larger errors occur, they are due to an underestimation of the storm intensity, rather than to biases in the boundary layer parameterizations.

Full access
Tom H. Durrant
,
Frank Woodcock
, and
Diana J. M. Greenslade

Abstract

The use of numerical guidance has become integral to the process of modern weather forecasting. Using various techniques, postprocessing of numerical model output has been shown to mitigate some of the deficiencies of these models, producing more accurate forecasts. The operational consensus forecast scheme uses past performance to bias-correct and combine numerical forecasts to produce an improved forecast at locations where recent observations are available. This technique was applied to forecasts of significant wave height (Hs ), peak period (Tp ), and 10-m wind speed (U 10) from 10 numerical wave models, at 14 buoy sites located around North America. Results show the best forecast is achieved with a weighted average of bias-corrected components for both Hs and Tp , while a weighted average of linear-corrected components gives the best results for U 10. For 24-h forecasts, improvements of 36%, 47%, and 31%, in root-mean-square-error values over the mean raw model components are achieved, or 14%, 22%, and 18% over the best individual model. Similar gains in forecast skill are retained out to 5 days. By reducing the number of models used in the construction of consensus forecasts, it is found that little forecast skill is gained beyond five or six model components, with the independence of these components, as well as individual component’s quality, being important considerations. It is noted that for Hs it is possible to beat the best individual model with a composite forecast of the worst four.

Full access
Tom H. Durrant
,
Diana J. M. Greenslade
,
Ian Simmonds
, and
Frank Woodcock

Abstract

This study examines the application of three different variations of linear-regression corrections to the surface marine winds from the Australian Bureau of Meteorology’s recently implemented operational atmospheric model. A simple correction over the entire domain is found to inadequately account for geographical variation in the wind bias. This is addressed by considering corrections that vary in space. Further, these spatially varying corrections are extended to vary in time. In an operational environment, the error characteristics of the wind forcing can be expected to change over time with the evolution of the atmospheric model. This in turn requires any applied correction to be monitored and maintained. Motivated by a desire to avoid this manual maintenance, a self-learning correction method is proposed whereby spatially and temporally varying corrections are calculated in real time from a moving window of historical comparisons between observations and preceding forecasts. This technique is shown to effectively remove both global and regionally varying wind speed biases.

Full access