Estimating in "Vane": A Quantitative Description of Wind Speed Overestimation by Human Observers Versus Instrument Measurements

Human wind reports are a vital supplement to the relatively sparse network of automated weather stations in the United States, especially for localized convective winds. The primary database of hazardous weather events in the United States, Storm Data, heavily relies upon human wind estimates in describing the high wind events that it records. Consequently, both high wind climatologies and efforts to calibrate high wind forecasting tools can be frustrated by the unknown accuracy of these estimates. While it is commonly assumed by many researchers and forecasters that humans over-estimate wind speed, this result has not been demonstrated by a robust scientific analysis. Accusations of human overestimation typically go uncited in research literature.

This study quantitatively assesses the plausibility of human wind estimates in Storm Data between 1996–2013 by comparing them to instrumentally observed wind speeds from the Global Historical Climatology Network (GHCN). Nonconvective wind events in areas of flat terrain within the continental United States served as the basis for this analysis due to the relative spatial homogeneity of wind fields in these meteorological and geographic settings. A gust factor (GF), defined here as the ratio of the daily max gust to the daily average wind, was calculated for both Storm Data and GHCN wind gust magnitudes. GFs resulting from human wind estimates were compared to the distribution of 6,801 GHCN GFs to judge their plausibility.

Human gust estimates were found to frequently result in implausibly large GFs located in the upper tail of the observed GHCN GF distribution. Stated plainly, humans reported exceptional wind gusts much more frequently than they were measured by nearby instruments. As a rule of thumb, humans tended to overestimate gusts by approximately one third. This means that for the average nonconvective wind event considered by this study, a human observer would perceive a 45.4-mph (20.3 ms-1) wind gust to satisfy the 58-mph (25.9 ms-1) high wind warning criteria. While this study was conducted using nonconvective wind events, the human biases that were quantified should remain relatively consistent in other meteorological regimes. We discuss the implications of this finding for researchers either directly or indirectly employing Storm Data wind reports – even if the reported magnitudes are ignored, because "severe" reports based on human observations may not have been at 58 mph. This study highlights the need to develop a formal procedure for incorporating human estimates in long-term climatological high wind databases.