Dr Paul Abernethy

Dr Paul Abernethy

Paul works to develop statistical techniques that can be used to help improve our forecast accuracy at specific locations. The resolution of the Unified Model places certain limitations on the accuracy of our forecasts, especially at locations that have characteristics that are not typical of the surrounding area. For example, the site could be located at the bottom of a deep valley, near the coast or on the side of a mountain. Forecasts for such sites will unavoidably contain systematic errors that are related to the site's location. Paul's work develops and applies statistical techniques that aim to minimise these systematic errors and thereby improve the accuracy of our forecasts at these locations.

Paul has applied these statistical techniques to various resolution configurations of the Unified Model, including the North Atlantic and European (NAE) 12km model and the United Kingdom 4km model (UK4). In order to move from grid scale forecasts to site specific forecasts, it is necessary to use downscaling algorithms which take into account such things as surface roughness, altitude and local topography. However, these downscaling algorithms introduce systematic errors to the forecast, and Paul has worked to determine which downscaling algorithm provides the best overall wind speed forecast when used in combination with advanced statistical filtering techniques. In addition, Paul has also shown that further improvements to the wind speed forecasts can be achieved by 'blending' our filtered UK4 and NAE wind speed forecasts to produce a weighted average of the forecasts which has lower error statistics, and improved skill, when compared to the non-blended forecasts.

Most recently, Paul has developed a simple algorithm that allows us to use our deterministic UK4 and NAE models to forecast the wind speed distribution for a given site. This algorithm uses the mean wind speed forecast and the maximum (gust) wind speed forecast for that site to calculate the probability that the wind speed will be within a certain range, and hence gives a better indication of the power output that can be expected from that site for a given forecast period.

Career background and experience

Paul has been a member of the Post Processing team since he joined the Met Office in 2008. Prior to joining the Met Office, Paul completed a PhD in Physics at the Department of Physics at the University of York. Prior to this, Paul received an MPhys in Physics and Physics with Astrophysics, also from the University of York.

"We're constantly aiming to significantly improve the accuracy of our forecasts and this also applies to forecasting wind speed. Turbine location is often at a site that's not characteristic of the general area and these are the errors that we are working towards removing from the final output."

Last updated: 15 April 2016

    Share this: