Projected changes in global terrestrial near-surface wind speed in 1.5 °C–4.0 °C global warming levels

The Max Planck Institute Earth System Model (MPI-ESM) is used in this study, which is the largest ensemble (100-member ensemble of simulations) of a single state-of-the-art comprehensive climate model currently available (https://esgf-data.dkrz.de/projects/mpi-ge/, last accessed on 25 June 2021). The performance of MPI-ESM LEs has been evaluated (Reichler et al 2012, Charlton-Perez et al 2013), and which has been extensively used to understand the effects of internal variability and external forcing (Maher et al 2018). The MPI-ESM is a fully coupled atmosphere–ocean model with a horizontal resolution of 1.9° × 1.9° and 47 vertical layers up to 0.01 hPa (Reick et al 2013). Individual members of the MPI-ESM only differ in their initial conditions (Bittner et al 2016), and the initialization branching time for each member from the preindustrial control run is introduced in Maher et al (2019). The historical simulations are integrated from 1850 to 2005, driven by observed radiative forcing. Future projections, not forecasts, are taken considering the most pessimistic future GHG emission rates, the RCP8.5 scenario, which represents that GHG emissions will lead to 8.5 W m⁻² radiative forcing (Costoya et al 2020). Furthermore, the strong warming magnitude cannot be reached with low emission scenario. Therefore, the future simulation from 2006 to 2099 with the RCP8.5 is used in this study (Monerie et al 2017, Huang et al 2020). In MPI-ESM, the land component is the JSBACH model including dynamic vegetation and land use transitions with the standard fire module (Maher et al 2019). The land use and cover change is considered in MPI-ESM LEs, and which includes the natural and anthropogenic land cover change (Reick et al 2013). The natural land cover change features two types of competition: between the classes of grasses and woody types controlled by disturbances (fire and windthrow) and within those vegetation classes between different plant functional types based on relative net primary productivity advantages (Reick et al 2013). The anthropogenic land cover change implements the land use transition approach by Hurtt et al (2006).

Three global reanalysis datasets are used to estimate the performance of MPI-ESM in simulating the NSWS. (a) The European Centre for Medium-Range Weather Forecasts fifth generation of atmospheric reanalysis (ERA5), (b) the National Centers for Environmental Prediction and National Center for Atmospheric Research reanalysis-1 (NCEP1), and (c) the Japanese 55 year reanalysis from Japan Meteorological Agency (JRA55) (table S1 available online at stacks.iop.org/ERL/16/114016/mmedia). These datasets hanvestigate the NSWS changes at global-scale (Ramon et al 2019, Fan et al 2021). In this study, monthly mean values of the MPI-ESM LEs and the reanalysis datasets are used.

To examine whether the MPI-ESM can simulate the terrestrial stilling, Global Surface Summary of the Day (GSOD) database is used (available at ftp://ftp.ncdc.noaa.gov/pub/data/gsod, last accessed on 25 June 2021). GSOD database consists of the United States Air Force DATSAV3 surface dataset and the Federal Climate Complex Integrated Surface Hourly dataset (Funk et al 2015). To obtain a high-quality wind record, Zeng et al (2019) have employed strict section criteria to avoid incomplete data series, and they concluded that the NSWS data at 1435 stations provide the most useful information. We utilize the data from those same 1435 stations in this study. The stations mainly locate in NH, so we compare the observation with MPI-ESM across NH to evaluate the performance of MPI-ESM in simulating the stilling.

The observed global mean near-surface air temperature (GMST) is about 1 °C warming relative to the preindustrial period; therefore, we define the 1.0 °C warming period in 1995–2004 in MPI-ESM as the current climate (baseline period). As such, the GMST increased by 1.5 °C, 2 °C, 3 °C, 4 °C compared with the preindustrial level in the models corresponding to the following four 10 year periods in MPI-ESM: 2015–2024 (1.50 °C), 2031–2040 (2.00 °C), 2055–2064 (2.99 °C), and 2077–2086 (4.02 °C). The GMST and NSWS changes are analyzed relative to the current climate. The GMST in the recent 30 year (1986–2015) increased by approximately 0.69 °C above the preindustrial level. To maintain the GMST increased by the same amount (0.69 °C), the model-simulated 30 year (1971–2000) mean values of NSWS in MPI-ESM LEs are compared with three reanalysis datasets (ERA5, NCEP1, and JRA55) during the period (1986–2015) when the GMST increased by the same amount (∼0.69 °C) above the preindustrial level.

To understand the causes of NSWS changes under different GWLs, the meridional temperature gradient (MTG) is calculated. Four latitude zones are defined: LZ1 (60° N to 90° N; 180° W to 180° E), LZ2 (0° to 60° N; 180° W to 180° E), LZ3 (45° S to 90° S; 180° W to 180° E), and LZ4 (0° to 45° S; 180° W to 180° E). The MTG is calculated using equation (1)

$$\begin{equation}\left\{ {\begin{array}{*{20}{c}} {{{\left( {{\text{MTG}}} \right)}_{{\text{NH}}}} = {{\left( {{\text{GMST}}} \right)}_{{\text{LZ1}}}} - {{\left( {{\text{GMST}}} \right)}_{{\text{LZ2}}}}} \\ {{{\left( {{\text{MTG}}} \right)}_{{\text{SH}}}} = {{\left( {{\text{GMST}}} \right)}_{{\text{LZ3}}}} - {{\left( {{\text{GMST}}} \right)}_{{\text{LZ}}4}}} \end{array}} \right..\end{equation} \tag{ 1 }$$

The trend is calculated using the least-square method, and the significance of the result is tested using the two-tailed Student's t-test. Reviewing previous studies (Chevuturi et al 2018, Cannon 2020), the globe is divided into eight regions: North America (25° N–72° N, 170° W–50° W), South America (60° S–20° N, 80° W–30° W), Europe (30° W–45° E, 30° N–75° N), Central Asia (30° N–65° N, 45° E–100° E), South Asia (5° N–35° N, 65° E–100° E), East Asia (10° N–79° N, 100° E–150° E), Africa (50° S–10° S, 110° E–160° E), and Australia (40° S–20° N, 20° W–55° E) (figure 1(a)).

Zoom In Zoom Out Reset image size

Spatial patterns of NSWS in MPI-ESM and reanalysis datasets are compared (figure 1). The large values of NSWS (>4.0 m s⁻¹) mainly locate in North America, Northern Africa, Asia, and Australia, and the small values of NSWS (<2.0 m s⁻¹) mainly locate in South America and Southern Africa (figure 1(a)). A similar spatial pattern of NSWS climatology is well reproduced in ERA5 (figure 1(b)) and NCEP1 (figure 1(c)). The wind speed difference between MPI-ESM and ERA5 is between ±1.0 m s⁻¹ in most regions of globe (figure S1(a)), and the MPI-ESM and NCEP1 mainly show positive bias but for the regions between 20° N and 20° S (figure S1(b)). Previous study has revealed that the NSWS climatology in ERA5 is closest to the observation due to the high spatiotemporal resolution in ERA5 compared to other reanalysis datasets (Fan et al 2021). This result indirectly demonstrate that the MPI-ESM LEs can capture the observed NSWS climatology. The mean NSWS in JRA55 is smaller than that in others (figure 1(d)), and the wind speed difference between MPI-ESM LEs and JRA55 exhibits positive value across globe (figure S1(c)). Previous studies have illustrated that the negative NSWS bias in JRA55 generates from the lowermost atmospheric level, which is placed too high over regions with trees, thus reducing the wind speed when interpolated from that level down to the altitude of 10 m, and are not fully corrected in the data assimilation process (Torralba et al 2017, Zhang et al 2020). Furthermore, JRA55 has a coarse resolution. These could explain why JRA55 has smaller NSWS compared to other reanalysis datasets.

Temporal changes of the NSWS in three reanalysis datasets mainly locate in the inter-model spreads of LEs, although the interannual changes in NSWS are not consistent (figure S2(a)). It is generally believed that the LEs can capture the climatology and dynamics of variables if the temporal series of reanalysis products can be included in the inter-model spreads of LEs (Li et al 2019). A significant decrease in NSWS is found in JRA55 (−0.021 m s⁻¹ decade⁻¹; p< 0.05), and this could be because the observed NSWS is assimilated in JRA55. According to our knowledge, only JRA55 assimilate observed NSWS over land into the reanalysis, but other reanalysis products assimilate winds over ocean (Fujiwara et al 2017, Torralba et al 2017, Zhang and Wang 2020). JRA55 simulates the long-term trend of NSWS could be better than other reanalysis products, but which shows largest absolute deviation of NSWS. Hence, it cannot be said that JRA55 simulates the global NSWS is better or worse than other reanalysis products. To confirm whether the stilling can be observed in LEs, the ensemble mean of members with negative trends in MPI-ESM is compared with the observation over the NH (figure S2(b)). The stilling can be observed in MPI-ESM LEs, although the stilling in MPI-ESM (−0.03 m s⁻¹ decade⁻¹; p< 0.01) is weaker than observation (−0.08 m s⁻¹ decade⁻¹; p< 0.01). In this study, we focus on the effects of the specific GWL on the long-term changes in NSWS when the effects of internal variability are excluded. The magnitude of trend in ensemble mean NSWS in MPI-ESM is smaller than observation, which is likely because the effects of internal variability are counteracted and the interannual fluctuations of NSWS are smoothed when the ensemble mean is carried out (Deser et al 2012, Wu et al 2018b, Shen et al 2021a).

Figure 2 shows the spatial patterns of annual mean NSWS changes with different GWLs. As the GMST increases by 1.5 °C–4 °C, the NSWS decreases over the NH mid-to-high latitudes and increases over the SH (except for Australia). The reduction in NSWS is more significant over North America and Eurasia and the increase is more pronounced over South America and Southern Africa. Relative to the current climate, under 1.5 °C, 2 °C, 3 °C, and 4 °C GWLs, the NSWS decreases by −0.015 ± 0.003, −0.016 ± 0.004, −0.029 ± 0.003, and −0.045 ± 0.004 m s⁻¹ over the globe, respectively; it decreases (increases) by −0.030 ± 0.0023 (+0.023 ± 0.005), −0.041 ± 0.005 (+0.052 ± 0.005), −0.070 ± 0.003 (+0.078 ± 0.005), and −0.095 ± 0.005 (+0.089 ± 0.007) m s⁻¹ over the NH (SH), respectively. With the same GWL, a substantial decrease in NSWS is detected in Asia and, with the strongest decrease occurs in South Asia (figure 3). Additionally, a substantial increase occurs in South America, especially for the 4 °C GWL where the future NSWS increases by +0.130 ± 0.010 m s⁻¹ relative to the current climate (figure 3). Consequently, the projected asymmetry changes in NSWS between two hemispheres could strengthen with the intensified GWL. Similar to the annual mean NSWS, the seasonal NSWSs also mainly decrease over the NH mid-to-high latitudes and increase over the SH; however, the seasonal NSWSs do not exhibit the decreasing trend over most regions of East Asia with the low GWL (e.g. 1.5 °C GWL) (figure S3). The temporal changes of the seasonal NSWS can also reveal the interhemispheric asymmetry of NSWS changes, although the seasonal differences of NSWS are evident and the interannual variations of NSWS are different among seasons (figure S4).

Zoom In Zoom Out Reset image size

Zoom In Zoom Out Reset image size

Figure 4 shows scatter plots between the NSWS and GMST changes. There is a linear relationship between the global NSWS decreases and GMST increases. The correlation coefficient is −0.91 (p< 0.01) (figure 4(a)). The NSWS and GMST are also negatively related over the NH (−0.98, p< 0.01). In contrast, these variables are positively related over the SH (0.94, p< 0.01). With 1.5 °C–2 °C (0.5 °C), 1.5 °C–3 °C (1.5 °C), and 1.5 °C–4 °C (2.5 °C) warming, the NSWS decreases −0.012 ± 0.004, −0.040 ± 0.004, and −0.066 ± 0.004 m s⁻¹ over the NH, respectively (figure 4(b)), and which increases +0.029 ± 0.006, +0.048 ± 0.004, and +0.069 ± 0.004 m s⁻¹ over the SH, respectively (figure 4(c)). The NSWS and GMST changes also exhibit negative correlations in most regions of NH and positive correlations in South America and Africa of SH (figure S5). These results confirm that decreases in global NSWS attribute to the intensified global warming occur mainly in the NH; whilst the NSWS over the SH could increase. There is low confidence in projected mean wind speed over Australia, with a medium confidence increase projected in northeastern Australia under high emission scenarios by the end of the 21st century (IPCC 2021). The monsoon is one of the most important climatic phenomena on Earth that can influence the wind changes (An et al 2015). The differential heating between land and ocean induced by monsoon causes the temperature gradient changes, and the latter can induce the changes in pressure gradient that is the driving force of wind speed changes (Wu et al 2018a). Noting that the projected Australian monsoon shows large uncertainty (IPCC 2021); therefore low confidence in projected NSWS changes over Australia may be linked to the large uncertainty in projected monsoon circulation changes.

Zoom In Zoom Out Reset image size

Probability density functions (PDFs) of NSWS under different GWLs are investigated (figures 4(d)–(f)). The NSWS PDFs clearly shift leftward but with little change in shape between the current climate and 4.0 °C warming, indicating a reduction in NSWS mean value but not the variability in NSWS (figure 4(d)). In terms of the NH, the NSWS PDFs also shift leftward but with little change in shape between the current climate and 4.0 °C warming; nevertheless, the NSWS over the NH is larger than global NSWS (figure 4(e)). With respect to SH, the NSWS PDFs shift rightward and the maximum probability density of NSWS reduces with the strengthened global warming, indicating an increase in NSWS mean value over the SH (figure 4(f)). Regional characteristics of NSWS PDFs show that the enhanced global warming causes the NSWS PDFs to move toward weak wind over North America, Europe, and Asia, especially for Central Asia, East Asia and South Asia, this characteristic is more pronounced. Nevertheless, the NSWS PDFs show opposite changes over South America and Africa, indicating that the intensified global warming is beneficial to the increase in NSWS over these two regions (figure S6).

It is acknowledged that the global warming is more pronounced over high latitudes than that over low latitudes (IPCC 2021). In this study, we also confirm that the increase in GMST is more significant over the NH mid-to-high latitudes than that in other regions under different GWLs based on MPI-ESM LEs (figure S7). It is worth noting that the increase in GMST over the SH, especially for the regions between 45 and 80° S, is less than 1.0 °C even where strong global warming occurs (figures S7(c) and (d)).

Temperature gradient is the primary driving force of NSWS (Wu et al 2018a). Global warming could induce the changes in temperature gradient (You et al 2010). To understand how enhanced warming influences the NSWS changes, the MTG is calculated. The MTG decreases over the NH and increases over the SH, at rates of −0.027 and +0.210 °C decade⁻¹, respectively, and that these characteristics are reproduced in the inter-model spreads of 100 members (figures 5(a) and (b)). The MTG and NSWS anomaly show positive correlation, with correlation coefficients of 0.88 (p< 0.01) over NH and 0.96 (p< 0.01) over SH (figure 5(c) and (d)). Quantitatively, each 0.5 °C decrease (rise) in MTG over NH (SH) causes a −0.20 ± 0.018 m s⁻¹ (+0.28 ± 0.001 m s⁻¹) decrease (increase) in NSWS over NH (SH). Briefly, the MTG decreases in NH and increases in SH attributes to unevenly global warming may be responsible for the NSWS decreases in NH and increases in SH.

Zoom In Zoom Out Reset image size

Changes in vertical atmospheric circulation could affect NSWS changes, so we detect the effects of different GWLs on the vertical atmospheric circulation. Three cells (Hadley, Ferrell, and Polar cells) are fairly reproduced with four GWLs (figure S8). However, with the enhancement of global warming, ascending flows are detected between 20 and 30° N and between 70 and 90° N, and descending flows are detected between 40 and 60° N (figure 6). Compared to the climatology of vertical atmospheric circulation (figure S8), the Hadley, Ferrell, and Polar cells are restrained due to the opposite vertical motions occur in the regions of descending flows in the Hadley cell and polar cell, and ascending flows in the Ferrell cell; as well as the inhibiting effects are more pronounced with the stronger GWL. Consequently, the strengthening in global warming may cause the weakening of the three cells over NH, and the latter result in a reduction in NSWS over NH. In SH, with the enhancement of global warming, descending flows are detected between 30 and 52° S and ascending flows are detected between 52 and 67° S. Consequently, the descending flows between 30 and 52° S triggered by strong GWL heighten the descending flows of the Hadley cell over the SH, while the location of descending flows in the Hadley cell moves slightly southward (Hu and Fu 2007). Consequently, the strengthened Hadley cell strengthens in SH due to the strengthening of global warming could induce the increase in NSWS over SH. Actually, these vertical cells influence historical wind speed changes have been discussed in the previous studies (Hu and Fu 2007, Deng et al 2021, Shen et al 2021a).

Zoom In Zoom Out Reset image size