资源环境科技发展态势分析平台

Reanalysis data underpin much research in atmospheric and related sciences. While most reanalysis only cover the last couple of decades, National Oceanic and Atmospheric Administration (20CR) and European Centre for Medium-Range Weather Forecasts (ERA20C and CERA20C) also developed reanalyses for the entire twentieth century that theoretically allow investigation of multidecadal variability. However, the approaches adopted to handle the massively evolving number of observations can cause spurious signals. Here we focus on wind speeds, as its assimilation is a key difference among these two products. We show that ERA20C and CERA20C feature significant trends in the North Atlantic and North Pacific wind speeds of up to 3m/s per century. We show that there is a good relation between the trends in the reanalysis and assimilated wind speeds. In contrast, 20CR and the European Centre for Medium-Range Weather Forecasts free model run ERA20CM do not show positive trends in the same regions. As a consequence, conclusions drawn from any single twentieth century reanalysis should be treated cautiously in particular in sectors with a strong wind dependency (e.g., wind energy).

Plain Language Summary Many areas of human activity are directly influenced by the climate, and an enhanced understanding of its variability is hence beneficial for the society. We need long-term climate data sets in order to quantify and understand climate variability better. As of today, there are two centers that provide gridded climate data sets for the last century (so called twentieth century reanalysis). Deriving these data sets is intricate because the number and quality of observations has changed dramatically during the period of interest. In our study, we show that the data sets disagree strongly with respect to long-term wind speed trends. As the climate system is highly coupled, other climatic variables are likely also affected. We analyze the underlying observational data, and we can show that the upward trends in one data set also exist in the observations. Furthermore, we can rule out that the model itself created the trends. By comparison with earlier studies, we argue that the trends are likely spurious (i.e., not real) but some uncertainty remains. We recommend that climate impact assessments should be based on data from both centers. In future research projects, attempts must be made to resolve the strong discrepancy between the data sets.