**Spatial effects of wind speed on SIF and GPP.**

The effects of wind speed on SIF and GPP were studied by using the standardized regression coefficient *K* and coefficient of determination *R*2 of a linear model with wind speed as the independent variable and SIF/GPP as the dependent variable (See Methods). Figure 1 (A, B) present the *K* spatial distributions for wind speed effects on SIF/GPP, respectively (The *R*2 distribution can be found in Figure S1 in Supplementary Information). A *K* larger (less) than zero indicates that wind speed has positive (negative) effect on SIF or GPP. A higher *R*2 indicates that wind speed can be used to predict more variations in SIF or GPP. Areas with an insignificant *K* (*p* > 0.05) were excluded. The results show that 61.2% of the areas for SIF and 44.6% of the areas for GPP were significantly affected by wind. The spatial distribution pattern of *K* (Fig. 1A, B) and *R*2 (Figure S1 A, B) are almost the same for SIF and GPP except for the region from 60°N to 90°N, where the effects of wind speed on GPP are insignificant but the effects on SIF are significant and positive. In the Amazon Rain Forest, Eurasia, and some coastal regions (Australia, South Africa, Oman, and India), the *K* value is greater than that in other regions for SIF and GPP (Fig. 1A, B). Regions of eastern United States (US), Brazilian Plateau, and some areas in Africa (Lunda Plateau, Katanga Plateau, and approximately the 10°N region) have lager *R*2 values than those in other regions for both SIF and GPP (Figure S1 A, B).

The areas with stronger positive (*K* being close to 1) wind speed effect on SIF and GPP are mainly located in the northern part of South America (Fig. 1A, B), where most of the vegetation types are broadleaf evergreen forests. The areas with stronger negative (*K* being close to -1) wind speed effect on SIF and GPP are mainly located in central Africa (Fig. 1A, B), where most of the vegetation types are grasslands. For the rest of areas, distributions of wind speed effect on SIF and GPP are scattered without an obvious pattern.

By analyzing the spatial distribution of wind speed effects on SIF and GPP, we found that the influence of wind speed on vegetation may be related to vegetation type, but different types of vegetation may not be present in all areas. Therefore, we further studied the influence of wind speed on different vegetations. Based on the MODIS data (MCD12C1), ten types of major vegetation were studied (evergreen needleleaf forests, evergreen broadleaf forest, deciduous needleleaf forests, deciduous broadleaf forests, mixed forests, closed shrublands, open shrublands, woody savannas, savannas, and grasslands) (Fig. 2A). Figure 2B shows the annual wind speed mean values (from 2001 to 2018) for the ten types of vegetation. The mean annual wind speeds in descending order for these vegetation types are: open shrublands, closed shrublands, grasslands, evergreen needleleaf forests, deciduous broadleaf forests, mixed forests, woody savannas, savannas, deciduous needleleaf forests, and evergreen broadleaf forests. Figure 2C shows the standardized regression coefficients *K* (regression models for SIF or GPP vs. wind speed) for the ten vegetation types. The averages of *K* for SIF in descending order for these vegetation types are: open shrublands, evergreen broadleaf forests, deciduous needleleaf forests, grasslands, closed shrublands, woody savannas, evergreen needleleaf forests, savannas, mixed forests, and deciduous broadleaf forests. The averages of *K* for GPP in descending order of magnitude for these vegetation types are: evergreen broadleaf forests, open shrublands, closed shrublands, deciduous needleleaf forests, grasslands, woody savannas, evergreen needleleaf forests, mixed forests, savannas, and deciduous broadleaf forests. Although, the orders for SIF and GPP are not exactly the same, they are very similar considering the large span of vegetation distribution. In general. Of these ten vegetation types, the proportions of wind speed that had a positive effect on SIF and GPP are all less than 50%; however, the proportion of wind speed that had a positive effect on SIF and GPP in evergreen broadleaf forests (SIF in open Shrublands) was the closest to 50%. SIF and GPP for most other vegetation types are under negative influence of wind speed (Fig. 2D), indicating winds being too strong for their growth need directly or indirectly. The specific spatial distribution of wind speed effects on SIF and GPP for different vegetations can be seen in Figure S4 and Figure S5, respectively. In north-central South America, wind speed positively influences the SIF and GPP of the local evergreen broadleaf forest overall. For most of the evergreen needleleaf forests, deciduous needleleaf forests, deciduous broadleaf forests, mixed forests, and open shrublands, the lower the wind speed, the higher the SIF and the GPP. In most grassland areas near the equator, wind speed has a strong negative effect on SIF and GPP.

GPP or SIF might change as a result of various reasons over 18 years. For further analysis of wind speed effects, it is meaningful to evaluate wind effect in the areas with increasing, decreasing, or no change in SIF/GPP over the 18 years individually. For GPP, the global areas are separated into three categories: areas with significant increase in GPP (AWSI), areas with significant decrease in GPP (AWSD), and areas without significant change in GPP (AWSC) (Figure S2, see Supplementary Information) according to GPP trend of variation (Eq. 1). The AWSI, the AWSD, and the AWSC account 28%, 3%, and 69% (*p* < 0.05) of the total areas, respectively. The color in Figs. 3A, 3B, and 3C represents the standardized regression coefficient *K* between wind speed and GPP. Area percentage for *K* < 0 accounts for 88.2%, 88.3%, and 89.0% in AWSI, AWSD and AWSC areas, respectively; area percentage for *K* > 0 accounts 11.8%, 11.7%, and 11.0% in AWSI, AWSD, and AWSC, respectively (Fig. 3D). Surprisingly, the area proportion with *K* > 0 or *K* < 0 is almost the same in AWSI, AWSD, and AWSC for GPP. Following the same procedures, the results for SIF can be computed, which also show that the area proportion with *K* > 0 (about 12%) or *K* < 0 (about 88%) is almost the same in AWSI, AWSD, and AWSC for SIF (Figure S3).

**Temporal effects of wind speed on SIF and GPP.**

In order to analyze the temporal effect of wind speeds on vegetation, we took the SIF and GPP data of 18 years (from 2001 to 2018) and divided them into 13 consecutive 6-year sliding windows in 1-year increment: the first period is from 2001 to 2006, and the second period is from 2002 to 2007, and so on (See Methods). In each of the six years, a regression equation between wind speed and SIF or GPP was established and the proportion of the areas with *K* > 0 was recorded. The percentage of the areas with *K* > 0 increases by nearly 2.8 times from 8.77–24.66% for SIF, while by about 2.5 times from 6.26–15.79% for GPP, which shows a successively increasing trend over time for SIF and GPP (*p* < 0.001) (Fig. 4A). The average of standardized regression coefficients *K* reduced by 60% from − 0.45 to -0.27 for SIF, and by 72% from − 0.50 to -0.36 for GPP, which also shows a successively increasing trend over time for SIF and GPP (*p* < 0.001) (Fig. 4B).

Similar to studying the spatial effect, we studied the change in wind speed effects on SIF and GPP of different vegetations from 2001 to 2018. The percentage of the areas with *K* > 0 increased by1.0% (1.0%), 1.2% (1.2%), 2.8% (2.6%), 0.3% (0.3%), 0.3% (0.3%), 3.9% (0.6%), 1.7% (1.6%), 1.7% (0.9%), and 1.7% (0.8%) in per 6 consecutive years for SIF (GPP) in evergreen needleleaf forests, evergreen broadleaf forests, deciduous needleleaf forests, deciduous broadleaf forests, mixed forests, open shrublands, woody savannas, savannas, and grasslands, respectively, which shows a successively increasing trend with time (*p* < 0.001); however, for the SIF and GPP of closed shrublands, the percentage of the areas with *K* > 0 did not change significantly (*p* > 0.5) (Fig. 5A and B); the average of standardized regression coefficient *K* increased by 0.02 (0.02), 0.01 (0.01), 0.03 (0.03), 0.01 (0.01), 0.01 (0.01), 0.01 (0.01), 0.03 (0.01), 0.02 (0.02), 0.02 (0.01), and 0.02 (0.01) per 6 consecutive years for SIF (GPP) in evergreen needleleaf forests, evergreen broadleaf forests, deciduous needleleaf forests, deciduous broadleaf forests, mixed forests, closed shrublands, open shrublands, woody savannas, savannas, and grasslands, respectively, which again shows a successively increasing trend with time (*p* < 0.05) (Fig. 5C and D).

**Optimal wind speed for the growth of each type of vegetation.**

Optimal wind speed for the growth of each type of vegetation was analyzed based on GIP and SIF. To do so, the 18 years of wind speed data were synthesized into annual mean wind speed (Fig. 6A) and then the annual mean wind speed data were further classified into 10 intervals by using the Jenks Natural Breaks Classification to maximize the difference between two neighboring intervals (See Methods), which tends to increase from low to high latitudes and from low to high altitudes. The 10 intervals are (0.59 to 1.51), (1.51 to 2.39), (2.39 to 3.01), (3.01 to 3.56), (3.56 to 4.10), (4.10 to 4.68), (4.68 to 5.35), (5.35 to 6.06), (6.06 to 7.19), and (7.19 to 11.25) m/s. Figure 6B shows the global mean wind speed distribution, which is close to a normal distribution.

For each type of vegetation, a map pixel was classified into one of the 10 wind speed intervals first. If a wind speed interval contains fewer than 5% of the total number of pixels, the interval would be ignored (See Methods, Figure S6 and Figure S7). Linear regression was performed between the monthly SIF (GPP) and monthly average wind speed for each of the pixel. The average *K* and the pixel percentage with *K* > 0 (*p* < 0.05) were computed for each of the 10 wind speed intervals for each of the vegetation types, and the results are shown in Fig. 6 and Figure S8, respectively.

As seen in Figs. 6C, D, E, F, G, H, I, J, K, and L, the effects of wind speed on SIF and GPP are similar in terms of average *K* among the different vegetation types. Evergreen needleleaf forests, evergreen broadleaf forests, deciduous needleleaf forests, deciduous broadleaf forests, and mixed forests belong to the forest system and their distributions are more concentrated in certain latitudinal ranges (Fig. 2A). It is relatively clear that the SIF and GPP of these five forest vegetation types have decreasing wind effect (reducing *K*) with increasing wind speed (Fig. 6C, D, E, F, and G). *K* is close to 0 in the (0.59, 1.51) m/s interval for evergreen broadleaf forests and in the (1.51, 2.39) m/s interval for deciduous needleleaf forests. The proportion of areas with positive impact of wind speed on SIF and GPP in evergreen needleleaf forests, deciduous broadleaf forests, and mixed forests is less than 10% (Figures S8 A, B, C, D, and E), but the percentage of areas with positive wind speed effects is close to 50% in the (0.59, 1.51) m/s interval for evergreen broadleaf forests and in the (1.51, 2.39) m/s interval for deciduous needleleaf forests, indicating that globally, based on SIF and GPP, evergreen broadleaf forests and deciduous needleleaf forests prefer higher mean annual wind speeds than the other three types of forest vegetation. Closed shrublands, open shrublands, woody savannas, savannas, and grasslands are globally dispersed in such a large latitudinal range that the optimal wind speed varies for them because of variations in other environmental conditions. According to Fig. 6A, wind speed increases from low to high latitudes and from low to high altitudes. Optimal *K* values for the growth of savannas and grasslands also increase with latitude or altitude (Fig. 6K and L). The effect of wind speed on SIF of open shrublands has a clear trend of changing from positive to negative with increasing latitude or altitude, while the trend of wind speed effect on GPP of open shrublands over latitude or altitude is not so obvious (Fig. 6I). The changes in *K* values for closed shrublands and woody savannas are not significant with increasing latitude or altitude (Fig. 6H, J).