Annual climatology of 10-m wind speed for (A) the preindustrial climate and (B) a simulation including turbine–atmosphere interactions for a globally homogeneous wind turbine density of one per 1 km².

KEE rate ignoring turbine–atmosphere interactions computed based on preindustrial near-surface wind speed climatology shown in Fig. S1A. For the diagnostic, a homogeneous global distribution of noninteractive turbines spaced 1 km² apart is assumed. Turbine specifications are identical to turbine settings prescribed in simulation with globally homogeneously spaced interactive turbines (i.e., turbines exerting atmospheric drag), for which the 10-m wind speed climatology is shown in Fig. S1B.

(A, C, E, G, I, K, and M) Mean near-surface KE dissipation and (B, D, F, H, J, L, and N) KEE for all wind farm simulations not shown in Fig. 3. (A–F) Atlantic open ocean wind farms with areas of 0.07, 0.21, and 0.67 Mkm² and (G–N) onshore wind farms in North America with areas of 0.1, 0.29, and 0.93 Mkm². Locations of wind farms are shown in Fig. 2.

Regional medians (•) and minimum–maximum ranges (lines) of annual mean KEE in (A) watts meter⁻² and (B) terawatts. Colors correspond to largest onshore (x-axis label lnd) and open ocean (x-axis label ocn) simulated wind farm domains shown in Fig. 2A. Open ocean and onshore wind areas vary slightly in latitude as explained in Methodology. These simulations have an effective turbine spacing of four times the turbine blade diameter (x-axis label 4D). Increasing the interturbine spacing to more realistic spacing of 10 times the turbine blade diameter (x-axis label 10D) reduces the mean extracted power by 31% over the ocean and 51% on land.

Time series of annual mean total near-surface dissipation of KE determined as the summed dissipation of KEE by the wind turbines and dissipation caused by surface drag. Reference climate time series (black line) shows surface dissipation only (KEE caused by wind turbines is 0 TW). Colors of wind farm simulations correspond to wind farm areas shown in Fig. 2.

(A) Seasonal variability for onshore wind farms in North America (Fig. 2A shows wind farm placement). Gray hatching indicates rate of KEE required to meet monthly mean electricity demand of the European Union (0.3–0.4 TW) scaled to wind farm size. (B) Interannual KEE variability determined over a 50-y analysis period for all wind farm simulations. On interannual timescales, the relative difference between minimal and maximal extraction rates of different years is between 30 and 35%, which is consistent with variability estimates for wind energy (36) and wind speed climatologies (37).

(I–III) Preindustrial climatology of 10-m wind speed (U_10m), surface temperature (Tsfc), and precipitation (P). Climatological absolute change of (A, D, G, J, M, P, S, and V) 10-m wind speed, (B, E, H, K, N, Q, T, and W) surface air temperature, and (C, F, I, L, O, R, U, and X) surface precipitation for all wind farm simulations. (A–L) Atlantic open ocean wind farms and (M–X) onshore wind farms in North America. Locations of wind farms are shown in Fig. 2, and wind farm areas are listed in Table S1.

Absolute difference in (A) geopotential height of the 950-hPa isobar (situated around 502 ± 100 m in the preindustrial climate throughout the geographic region), (B) sea ice fraction, (C) total cloud cover, and (D) net shortwave radiation at the surface (defined as positive down; watts meter⁻²) between the largest open ocean wind farm simulation and the preindustrial climate. Large-scale dynamical effects resulting from the presence of giant wind farms have previously been addressed in ref. 38, where it was shown that wind farms of comparable size may induce Rossby waves in the large-scale flow in highly idealized simulations. In our case, the large surface drag in wind farms with an area exceeding 0.1 Mkm² partially deviates the horizontal flow around the wind farm, inducing a closed cyclonic tendency to the wind field north of the wind farm and an anticyclonic tendency to the south of the wind farm in the climatological mean. In the Arctic, this leads to a dynamical sea ice feedback during the winter months, where the sea ice edge is pushed farther south.

(A and B) Same as in Fig. 2 but for May to September averages instead of annual means. (C) Interannual variability for the May to September period only. Colors of wind farm simulations correspond to wind farm areas shown in Fig. 2. Exact values for KEE are summarized in Table S2.

Median and interquartile percentiles of KEE rate (watts meter⁻²) against net surface heat flux (watts meter⁻²). Percentiles were computed for the global wind farm simulation over (A) the Northern Hemisphere midlatitudes (30° N to 66° N), (B) the subtropics and the tropics (30° S to 30° N), and (C) the Southern Hemisphere midlatitudes (30° S to 66° S). The net heat flux is defined as positive upward. Percentiles for land and ocean points were determined separately and are shown in green and blue, respectively.