Can Surface Roughness Changes Explain Terrestrial Wind Stilling?

Surface wind speeds have exhibited statistically significant declining trends (“stilling”) for 30 – 50 years over much of the land surface, with notable increasing trends over the oceans, at high latitudes and at altitudes above the lower boundary layer. These trends are not captured in global climate models or in retrospective analyses (reanalyses) over the same period, with most exhibiting slightly positive or negative trends that are often statistically insignificant. This is particularly important as models are the only means by which to obtain data in regions not well sampled by stations or for analyses of future wind variability under climate change. Various reasons for this discrepancy have been postulated throughout the literature. Climate changes and variability have been cited, though it remains unclear why reanalyses, in particular, would not register a strong climate signal in the wind records, given that they are driven by large sets of observations. Inhomogeneities in the station wind records have also been cited as possible sources of the discrepancy, though previous work has reached conflicting conclusions as to the real impact of changes in station location, anemometer height and instrumentation on wind trends. The other possibility noted in studies of several wind records around the world, including important regions for wind energy production, is the influence of surface roughness shifts induced through land cover change. Work using models and extrapolation of roughness changes from remote sensing data indicated that roughness changes could account for 25 – 60 percent of the wind speed declines, though no long-term datasets of concurrent land cover changes were available to corroborate these findings. The study done by Meiyappan and Jain (2012), which recreated annual land cover types for the globe at 0.5° resolution over the industrial period, provides a useful dataset for this purpose. Using the estimates of 28 land cover types and assigning them representative roughness values in three distinct reconstructions, three estimates of roughness change for the period 1975 – 2005 were obtained by dividing the values for 2005 by those of 1975. This roughness factor was then used in an empirical equation taken from Vautard et al. (2010) to calculate the impact on wind speeds over the 30 year period and compared to regional wind declines documented in Bichet et al. (2012). It was found that stilling in response to roughness changes could account for, at maximum, 5% of the trend when averaged globally, with regional values that ranged from 0.4 – 11%. Changes in roughness varied by reconstruction and region, though some consistent patterns emerged in regions such as Europe, where roughness increased, the northern edges of the boreal forests where roughness declined, and in sparsely populated areas such as deserts and ice sheets which showed little change. To assess the importance of the uneven distribution of stations, which tend to be biased towards areas with significant human impact on land cover, grid cells were also averaged selectively based on the locations of Integrated Surface Database (ISD) sites at which wind had been measured within period. This increased the estimated impact of roughness on wind speeds to about 22%, suggesting that the changes in wind speeds due to roughness could be more pronounced in wind records than the regionally averaged results would indicate. Further, the influence of station inhomogeneities cannot be ruled out. Wind observations taken from the Upper Midwest for the period 1979 – 2016, when corrected for important station moves in the 1990s and instrumentation changes in the 2000s, exhibit significant reductions in absolute trends, for the region as whole, from -0.18 m/s/decade to -0.05 m/s/decade. The change in trend increases the estimated impact of roughness changes from 2% to 9% when regionally averaged and from 7% to 34% when the station locations are selectively analyzed, fitting the same pattern as the global analysis. Thus, while initial work indicates that roughness changes explain a very modest amount of the wind stilling for the regions analyzed, there is reason to suspect that data inhomogeneity may play an important role in explaining the seeming inconsistencies between climate models, reconstructed roughness changes, and observational records. Future research could focus on bias-correcting wind records for more regions and improving land cover change datasets to narrow the considerable uncertainties in this work.