Figure 2.38 Trends in surface wind speed for 1988–2010. Shown in the top row are data sets based on the satellite wind observations: (a) Cross-Calibrated Multi-Platform wind product (CCMP; Atlas et al., 2011); (b) wind speed from the Objectively Analyzed Air-Sea Heat Fluxes data set, release 3 (OAFlux); (c) Blended Sea Winds (BSW; Zhang et al., 2006); in the middle row are data sets based on surface observations: (d) ERA-Interim; (e) NCEP-NCAR, v.1 (NNR); (f) 20th Century Reanalysis (20CR, Compo et al., 2011), and in the bottom row are surface wind speeds from atmospheric reanalyses: (g) wind speed from the Surface Flux Data set, v.2, from NOC, Southampton, UK (Berry and Kent, 2009); (h) Wave- and Anemometer-based Sea Surface Wind (WASWind; Tokinaga and Xie, 2011a)); and (i) Surface Winds on the Land (Vautard et al., 2010). Wind speeds correspond to 10 m heights in all products. Land station winds (panel f) are also for 10 m (but anemometer height is not always reported) except for the Australian data where they correspond to 2 m height. To improve readability of plots, all data sets (including land station data) were averaged to the 4° × 4° uniform longitude-latitude grid. Trends were computed for the annually averaged timeseries of 4° × 4° cells. For all data sets except land station data, an annual mean was considered available only if monthly means for no less than eight months were available in that calendar year. Trend values were computed only if no less than 17 years had values and at least 1 year was available among the first and last 3 years of the period. White areas indicate incomplete or missing data. Black plus signs (+) indicate grid boxes where trends are significant (i.e., a trend of zero lies outside the 90% confidence interval).
AR4 concluded that mid-latitude westerly winds have generally increased in both hemispheres. Because of shortcomings in the observations, SREX stated that confidence in surface wind trends is low. Further studies assessed here confirm this assessment.
Surface wind measurements over land and ocean are based on largely separate observing systems. Early marine observations were based on ship speed through the water or sails carried or on visual estimates of sea state converted to the wind speed using the Beaufort scale. Anemometer measurements were introduced starting in the 1950s. The transition from Beaufort to measured winds introduced a spurious trend, compounded by an increase in mean anemometer height over time (Kent et al., 2007; Thomas et al., 2008). ICOADS release 2.5 (Woodruff et al., 2011) contains information on measurement methods and wind measurement heights, permitting adjustment for these effects.
The ICOADS-based data set WASWind (1950–2010; Tokinaga and Xie, 2011a) and the interpolated product NOCS v.2.0 (1973–present; Berry and Kent, 2011) include such corrections, among other improvements. Marine surface winds are also measured from space using various microwave range instruments: scatterometers and synthetic aperture radars retrieve wind vectors, while altimeters and passive radiometers measure wind speed only (Bourassa et al., 2010). The latter type provides the longest continuous record, starting in July 1987. Satellite-based interpolated marine surface wind data sets use objective analysis methods to blend together data from different satellites and atmospheric reanalyses. The latter provide wind directions as in Blended Sea Winds (BSW; Zhang et al., 2006), or background fields as in Cross-Calibrated Multi-Platform winds (CCMP; Atlas et al., 2011) and OAFlux (Yu and Weller, 2007). CCMP uses additional dynamical constraints, in situ data and a recently homogenized data set of SSM/I observations (Wentz et al., 2007), among other satellite sources. Figure 2.38 compares 1988–2010 linear trends in surface wind speeds from interpolated data sets based on satellite data, from interpolated and non-interpolated data sets based on in situ data, and from atmospheric reanalyses. Note that these trends over a 23-year-long period primarily reflect decadal variability in winds, rather than long-term climate change (Box 2.2). Kent et al. (2012) recently intercompared several of these data sets and found large differences. The differences in trend patterns in Figure 2.38 are large as well. Nevertheless, some statistically significant features are present in most data sets, including a pattern of positive and negative trend bands across the North Atlantic Ocean (Section 2.7.6.2.) and positive trends along the west coast of North America. Strengthening of the Southern Ocean winds, consistent with the increasing trend in the SAM (Section 2.7.8) and with the observed changes in wind stress fields described in Section 3.4.4, can be seen in satellite-based analyses and atmospheric reanalyses in Figure 2.38. Alternating Southern Ocean trend signs in the NOCS v.2.0 panel are due to interpolation of very sparse in situ data (cf. the panel for the uninterpolated WASWind product).
Surface winds over land have been measured with anemometers on a global scale for decades, but until recently the data have been rarely used for trend analysis. Global data sets lack important meta information on instrumentation and siting (McVicar et al., 2012). Long, homogenized instrumental records are rare (e.g., Usbeck et al., 2010; Wan et al., 2010). Moreover, wind speed trends are sensitive to the anemometer height (Troccoli et al., 2012). Winds near the surface can be derived from reanalysis products (Box 2.3), but discrepancies are found when comparing trends therein with trends for land stations (Smits et al., 2005; McVicar et al., 2008).
Over land, a weakening of seasonal and annual mean as well as maximum winds is reported for many regions from around the 1960s or 1970s to the early 2000s (a detailed review is given in McVicar et al. (2012)), including China and the Tibetan Plateau (Xu et al., 2006b; Guo et al., 2010) (but levelling off since 2000; Lin et al., 2012), Western and southern Europe (e.g., Earl et al., 2013), much of the USA (Pryor et al., 2007), Australia (McVicar et al., 2008) and southern and western Canada (Wan et al., 2010). Increasing wind speeds were found at high latitudes in both hemispheres, namely in Alaska from 1921 to 2001 (Lynch et al., 2004), in the central Canadian Arctic and Yukon from the 1950 to the 2000s (Wan et al., 2010) and in coastal Antarctica over the second half of the 20th century (Turner et al., 2005). A global review of 148 studies showed that near-surface terrestrial wind speeds are declining in the Tropics and the mid-latitudes of both hemispheres at a rate of −0.14 m s−1 per decade (McVicar et al., 2012). Vautard et al. (2010), analysing a global land surface wind data set from 1979 to 2008, found negative trends on the order of –0.1 m s–1 per decade over large portions of NH land areas. The wind speed trend pattern over land inferred from their data (1988–2010, Figure 2.38) has many points with magnitudes much larger than those in the reanalysis products, which appear to underestimate systematically the wind speed over land, as well as in coastal regions (Kent et al., 2012).
In summary, confidence is low in changes in surface wind speed over the land and over the oceans owing to remaining uncertainties in data sets and measures used.