Posters

Abstract

Wind resources show different degrees of variability in space and time. In general, the compounding of the climate change signal with high spatial and temporal wind variability can lead to large uncertainties in the projected impacts of climate change on wind resources.  One question often asked is which factors control or contribute to past variability, and how this could be taken into account in the EYA. This question cannot be answered by just focusing on the more recent 20 years, the time frame typically considered in these assessments. However, this focus on the most recent data is often justified by the limitations of older datasets.   How can we use these data if they actually increase calculated uncertainty? How much confidence can we have in wind speed data sets back to the 1960s? Are the older data sets less certain and does this affect how we should use them? For wind resource assessment, long and accurate time series with high sampling frequency are crucial to capture the short-term and multi-decadal variability. In reality, however, there is rarely a case where all these aspects are satisfied. Dedicated in-situ observations, e.g. with masts, provide high temporal resolution, representative for the specific location and they are typically extended with reanalysis data. In this context questions arise about data accuracy and reliability. ERA5 is one of the most frequently used reanalysis data sets due to its global coverage and its state-of-the-art spatial and temporal resolution. In addition, ERA5 provides a multi decades long record back until the mid of the 20th century. Here, we explore the different layers of the involved uncertainties specifically for wind speed, and for which use cases they are more or less relevant.  The accuracy of ERA5, and reanalysis data in general, crucially depends on the amount and quality of assimilated observations. Thus, we examine the intrinsic uncertainty of the reanalysis due to the availability of historical observations for the ERA5 data assimilation which can be assessed with the ERA5 ensemble spread. This measure can depend not only on the amount of data assimilated, but also on the exact variables that are available for assimilation at a certain location and time. For instance, near-surface wind is only taken into account over seas. Land areas are excluded.  We show how the ERA5 ensemble spread varies with time and space and identify potential regions and time periods of reduced reliability. Specifically with the decrease of available observations further back in time in mind, the results can provide guidance for where and when ERA5 is more or less valid. Together with a collection of other sources for systematic errors, which are not taken into account by the ERA5 ensemble spread, we provide a set of tools that enables an informed decision about the suitability of ERA5 data for wind resource assessments that take into account multi-decadal variability, depending on the location and the time scale.