Over recent years re-analysis wind data has gone from being a new-fangled curiosity to the current state of the art in long-term wind resource studies. But this new data source is causing analysts to cut corners drawn in by the potential, not the cold hard reality of this exciting source.

For those not familiar with wind resource assessment processes – typically an understanding of the wind speed at any location is gained through short-term measurements (a period of one to two years). In order to understand likely future production these short-term readings are adjusted to the long-term using a reference source.

Traditional reference sources consist of a short height mast maintained by a national meteorological service measuring wind speed and direction. In recent years, re-analysis data from global climate models, historically used for climate modelling and research, has been explored as an alternative. Re-analysis data uses inputs from many different sources including satellites, weather balloons and ground station across the globe over many years.

On the face of it re-analysis data seems like the perfect fit for long-term wind resource assessment. The data is global, with high spatial and temporal resolution. Even better, it covers a time period much greater than all but the most robust ground measurement station. Every few months a new source becomes available offering better correlations, the latest being a step change improvement from ERA‑5. So that means we can consign the hours of screening metadata from anemometers mounted on a 10-metre mast maintained by the local meteorological service to history, right?

Wrong, it’s not that simple. Re-analysis data is a wonderful resource and one that has huge potential; but many questions remain unanswered and many of the answers found when you dig a little deeper raise serious concerns for the wind industry.

We’re seeing ever improving correlations due to new models (such as the recent release of ERA-5 and the development of the MERRA-2 dataset), but these improved models still don’t address one of the critical issues – consistency. Are the long-term trends in re-analysis models representative – or are they artefacts of the inputs or the model? With the millions of inputs, we can’t perform a bottom-up validation of re-analysis, which we can for the humble met station, so we need to use new techniques to validate and verify the outputs.

Practice often does not reflect the science. Analysts are imbuing re-analysis data sources with qualities that are not only unproven but in many cases have been demonstrated to be false.

This work has been ongoing within the industry for several years and the results clearly demonstrate that whilst re-analysis has potential to improve long-term wind estimates in many cases, it contains inconsistencies, it doesn’t accurately represent inter-annual variation in wind speed and the performance varies dramatically across the globe. In some regions these new sources are even producing worse long-term estimates than their predecessors.

Practice often does not reflect the science. Analysts are imbuing reanalysis data sources with qualities that are not only unproven but in many cases have been demonstrated to be false. It is not good enough to assume a reanalysis source is consistent and it is not good enough to only examine one or two sources of long-term data. If re-analysis data is to be used it has to be rigorously proven to be representative – the same as is required for the met station.

We’ve seen in the past the impact that errors in non-representative long-term reference sources can have on energy yield predictions. High wind speed years in the UK in the mid-1990s biased energy yield predictions around this period resulting in many projects being over-valued and undermined investor trust in the years that followed. Long-term wind speed is the cornerstone of any wind resource or energy yield prediction and drives the revenue of a project – so we need to get it right.

So here’s the challenge for you and me: we need to firstly fully validate the data that we have available and understand what we can use it for, and how good is it. Secondly work together with the climate modelling industry to improve on the products we use so that we can fully realise the potential of reanalysis. The future for re-analysis data is bright and it is undoubtedly the future – but we’re not there yet.

David Pullinger is technical lead – Energy Resource Services at Lloyd’s Register

This is part of a series of blog posts from RenewableUK in the run-up to Global Offshore Wind 2018. To find out more about progress on validation studies of reanalysis and mesoscale models head into session B8: Resource & Assessment 2:15pm on 20 June.