Widespread lagged and cumulative effects of wind on global vegetation productivity

Wind speed has undergone substantial changes over recent decades, significantly affecting vegetation growth. However, the temporal effect of wind on vegetation productivity remains poorly understood, limiting accurate assessments of wind-vegetation interactions. This study explored the lagged and cumulative effects of wind on vegetation productivity from 2000 to 2024 using multiple datasets. The results showed that wind exhibited significant temporal effects on approximately 37% of global vegetated areas, with both lagged and cumulative effects mainly occurring within a 1-month time scale. Overall, the time scales of lagged effects were mainly associated with changes in precipitation, whereas those of cumulative effects were primarily associated with changes in air temperature. In addition, changes in the temporal effect of wind were strongly influenced by soil and topographic factors, particularly soil organic carbon and elevation. These findings highlight the importance of incorporating wind-related temporal effects into terrestrial carbon models to improve our understanding of vegetation productivity dynamics under accelerating climate change.

Wind speed has experienced substantial changes at regional and global scales in recent decades (Dunn et al., 2024), with significant implications for ecosystem structure and function (Rey et al., 2012; Dong et al., 2023). As a key component of the climate system, wind can regulate vegetation productivity via multiple processes, including evapotranspiration (McVicar et al., 2012), gas exchange (Wu et al., 2018), and the transport of heat and moisture (Wu et al., 2025), with impacts in some cases comparable to those of temperature and water availability (Wu et al., 2021). Moreover, wind-driven mechanical stress influences vegetation morphology, where strong wind events can induce physical damage and increase vegetation mortality (Gardiner et al., 2016). Despite these recognized mechanisms, the temporal dynamics of wind-vegetation interactions across different regions remain largely unclear, which increases uncertainty in predicting terrestrial carbon sink responses to climate change.

It is widely recognized that vegetation productivity responses to climate change typically exhibit pronounced lagged and cumulative effects (Ding et al., 2020; Wei et al., 2022). The lagged effect refers to the influence of climatic conditions in preceding months on vegetation productivity (Wu et al., 2015; Peng et al., 2019), whereas cumulative effects represent the integrated influence of climatic conditions over multiple antecedent months, including the current month (Vicente-Serrano et al., 2013; Wen et al., 2019). However, most studies have mainly examined hydrothermal variables (i.e., temperature, precipitation, and solar radiation), with relatively limited attention given to the temporal effects of wind. Given that wind can modulate both land-atmosphere interactions and physiological processes, its lagged and cumulative effects on vegetation productivity warrant investigation.

Here, we investigated the temporal effect of wind on global vegetation productivity using long-term multisource datasets. Specifically, this study aims to (1) quantify lagged and cumulative time scales of wind affecting vegetation productivity, (2) examine differences in these temporal effects among vegetation types and hydrothermal conditions, and (3) identify the relative importance of biotic and abiotic drivers of these temporal effects. These findings are expected to advance the representation of wind-driven processes in climate models and improve predictions for vegetation responses under future climate scenarios.

Datasets

To explore vegetation productivity-wind relationships, long-term vegetation, climate, and relevant auxiliary data were collected. Gross primary productivity (GPP) was employed as an indicator for vegetation productivity and obtained from the 8-day GOFGPP dataset at a grid size of 0.05° (Li and Xiao, 2019). This GPP product has shown good reliability compared with ground-based observations across various ecosystems (Liu et al., 2023). The monthly mean wind speed, total precipitation, and mean air temperature data were derived from the ERA5 datasets at a grid size of 0.25°, widely used in ecological studies (Hersbach et al., 2020). The auxiliary data included digital elevation data from GETED2010 (Danielson and Gesch, 2011), vegetation type data from the ESA CCI product (ESA, 2017), soil property data (soil bulk density, soil clay content, and soil organic carbon) from the SoilGrids product (Hengl et al., 2017), biodiversity data from the number of plant species map (Kier et al., 2005), and human activity impact data from the GHM map (Kennedy et al., 2019).

Moreover, we retained pixels with an unchanged vegetation type and reclassified vegetation types into six major categories: grassland, shrubland, forest (needleleaf, broadleaf, mixed forest), wetland, cropland, and sparse vegetation. All datasets spanning from 2000 to 2024 were used in this study and aggregated to a uniform spatial grid of 0.25° before analysis.

Quantification of lagged and cumulative effects of wind on vegetation productivity

A linear regression approach was applied to quantify lagged and cumulative responses of vegetation productivity to wind. Given that the temporal effects of climate factors generally are shorter than a quarter (Wu et al., 2015; Ding et al., 2020), the analysis of wind-induced lagged and cumulative effects was restricted to a maximum time scale of 3 months. The relationship between wind and vegetation productivity can be calculated using the following equation (Equation 1): where m and n denote the fitted regression coefficients. The values of j and k range from 0 to 3, and their combinations defined four scenarios capturing different temporal effect patterns: (1) when j = 0, k = 0: no temporal effect; (2) when j = 0 with k ranging from 0 to 3: cumulative effects only; (3) when k = 0 with j ranging from 0 to 3, lagged effects only; and (4) if j and k both ranging from 0 to 3, both lagged and cumulative effects. Temporal effects were assessed within a 3-month window, with j + k ≤ 3 when both lagged and cumulative effects were considered.

This analysis focused on the growing season, corresponding to months with the mean air temperature exceeding 0 °C and normalized difference vegetation index values exceeding 0.2. The optimal temporal effects were selected as the model with the highest determination coefficient (R 2) with statistical significance (p < 0.05), corresponding to different combinations of lag (j) and cumulative (k) months. Before analyses, the GPP and wind speed time series were converted to anomalies to minimize the influence of their strong seasonal variability.

Assessment of drivers for lagged and cumulative effects of wind on vegetation productivity

We used the random forest (RF) model to assess the importance of factors influencing wind-induced lagged and cumulative effects on vegetation productivity. This data-based machine learning approach employs a bootstrap bagging method, offering robust predictive capability and enabling evaluation of feature importance (Breiman, 2001). Four types of drivers were considered: climate, soil, topography, and other relevant factors. Specifically, the annual mean total precipitation, air temperature, solar radiation, and aridity index were calculated as key climatic limiting factors. Topography factors, including elevation, slope, and aspect, play an important role in shaping the spatial patterns of climate variables and vegetation. Soil factors, such as soil organic carbon, soil clay content, and soil bulk density, were included due to their strong relationships with vegetation photosynthesis and the distribution of soil resources. Other factors closely associated with wind speed changes and vegetation productivity, including vegetation types, human activity influence, and biodiversity, were incorporated. In addition, a standardization method (Z-score) was applied to all variables to eliminate scale differences before analysis.

Although RF is relatively robust to multicollinearity, variables with high collinearity (i.e., VIF >5) were excluded based on the variance inflation factor (VIF) to avoid bias in feature importance. To analyze the time scales of wind-induced lagged and cumulative effects, two RF classification models were developed separately. The lagged and cumulative time scales were treated as response variables, while the four groups of factors served as predictors. The optimal hyperparameters, including the number of trees, maximum number of features, and maximum tree depth, were tuned using 80% of the total samples for model training, while the remaining 20% were reserved for independent testing. The models achieved moderate to high predictive performance in estimating the optimal timescales of the lagged and cumulative effects, with overall accuracies of 0.70 and 0.70, weighted-average F1 scores of 0.68 and 0.69, and Kappa coefficients of 0.62 and 0.60, respectively.

Spatial patterns of the temporal effect of wind on global vegetation productivity

We first quantified the time scales of wind-induced lagged and cumulative effects and found strong spatial heterogeneity across global vegetated areas (Figure 1). Temporal effects of wind were observed in approximately 37% of the study area (Figure 1c). Specifically, the lagged effect of wind primarily occurred at short time scales (≤1 month), with a global mean lag of 0.74 ± 0.96 months (Figure 1a). The largest proportion of areas exhibited no lag (23%), followed by 1 month (11%), both of which were widely distributed across the globe. A 2-month lag occurred in 5% of areas, mainly observed in Alaska, southwestern North America, Europe, southern Africa, and southern South America. The smallest proportion was observed for the 3-month lag (3%), primarily occurring in northwestern North America, northern China, and western Australia.

From the perspective of the cumulative effect of wind, the time scale distribution was similar to that of the lagged effect, primarily concentrated within 1 month (Figure 1b). At the global scale, the mean cumulative time of wind was 0.89 ± 0.99 months. Cumulative periods of 0–1 month accounted for 31% of global vegetated areas and were widely distributed globally. This was followed by a 2-month time scale, mainly occurring in northwestern North America, northern China, and eastern Russia. The smallest proportion corresponded to a 3-month time scale, mainly observed in southern North America, northwestern India, eastern Brazil, and southeastern Australia.

Temporal effects of wind on vegetation productivity under different vegetation types and hydrothermal conditions

Vegetation productivity showed markedly different lagged and cumulative responses to wind across vegetation types (Table 1). Overall, wind impacts were characterized by short temporal delays, with a mean lagged time of less than 1 month. Forests showed the strongest lagged response (0.82 ± 1.00 months), while grasslands and sparse vegetation showed weaker lag effects, and croplands and wetlands showed intermediate responses. In contrast, cumulative effects were more pronounced in shrublands (1.05 ± 1.04 months) and moderate in grasslands and croplands, while forests and wetlands showed relatively weaker cumulative responses.

We found a clustering effect of wind on vegetation productivity over hydrothermal gradients (Figure 2). For forest, longer lag times were mainly associated with regions where mean annual air temperature ranged from 4 °C to 20 °C. In contrast, for grassland and cropland, a longer lagged time was mainly found in regions where the annual total precipitation exceeded 400 mm. Moreover, precipitation plays a more important role than air temperature in regulating wind-induced lagged effects. Except for sparse vegetation, the lagged time of wind increased with increasing precipitation, with a rate of approximately 0.0002 (p < 0.05). The cumulative effect of wind also varied across different hydrothermal conditions (Figure 2). In contrast to the lagged response, it exhibited no clear dependence on the precipitation gradient. Instead, the cumulative time generally increased with increasing air temperature for most vegetation types, indicating a stronger wind effect under warmer conditions.

Drivers of the temporal effect of wind on global vegetation productivity

Identifying environmental drivers for lagged and cumulative time scales of wind is crucial for understanding wind-vegetation interactions. Generally, climate factors contributed the largest proportion (mean of ∼31%) to the temporal effects of wind on vegetation productivity (Figure 3), followed by soil factors (∼27%), topography factors (∼23%), and other related factors (∼18%). More specifically, air temperature was the most important climate variable for both lagged and cumulative effects, followed by the aridity index and precipitation. Among soil property factors, soil organic carbon was the most important for the temporal effects of wind on vegetation productivity. For topography factors, elevation was the dominant contributor, followed by slope and aspect. In comparison, biodiversity and vegetation type exhibited relatively low impacts.

In this study, we found that wind influenced vegetation productivity not only instantaneously but also had significant temporal effects. Lagged and cumulative effects of wind were mainly observed at short time scales (within 1 month), which were similar to temporal effects of air temperature and precipitation (Ding et al., 2020). The predominance of short-term temporal effects indicated that vegetation responses to wind are closely linked to rapid adjustments in physiological processes. Wind can directly modulate canopy conductance, evapotranspiration, and turbulent exchange (Wu et al., 2025), thereby influencing plant water use and photosynthetic activity within relatively short periods. Incorporating wind-induced temporal effects can improve the evaluation of climate-vegetation productivity interactions.

Moreover, the widespread yet spatially heterogeneous distribution of these temporal effects suggested that wind-driven processes are not universally uniform but are strongly conditioned by the regional environment (Figure 2). Regions characterized by low precipitation generally exhibited shorter lagged effects, whereas the time scale of cumulative effects increased with rising air temperature for most vegetation types. This divergence suggests that water availability and thermal conditions may influence different aspects of vegetation responses to wind. Greater precipitation generally maintains higher soil-moisture availability and plant water status, which may increase the capacity of vegetation to buffer wind-induced increases in boundary-layer conductance and transpiration. Consequently, wind-induced water loss may take longer to cause substantial changes in plant water status and stomatal regulation, resulting in a longer lagged response. In contrast, in water-limited regions, lower soil moisture availability reduces this buffering capacity, allowing wind-enhanced transpiration to more rapidly increase water stress and affect photosynthetic activity (Schymanski and Or, 2015), leading to shorter lagged time.

In contrast, the cumulative effect of wind is largely regulated by temperature, which controls phenological and photosynthetic processes that rely on sustained thermal conditions rather than short-term fluctuations (Piao et al., 2020). Higher temperature increases evaporative demand and leaf-to-air vapor pressure differences, potentially amplifying the effects of wind on boundary-layer conductance, transpiration, and leaf energy balance. Over multiple days, these effects may accumulate through progressive soil-moisture depletion, changes in plant water status, and adjustments in photosynthetic activity, resulting in longer cumulative timescales. Temperature may therefore be more closely associated with the persistence and accumulation of wind effects than with the initial delay before a vegetation response becomes detectable.

The contribution analysis also indicated that climate variables such as air temperature and precipitation dominated temporal effects of wind. This finding is consistent with previous studies highlighting the coupling between atmospheric demand and water availability in controlling vegetation productivity dynamics (Sulman et al., 2016; You et al., 2020). Differences among vegetation types may reflect variation in canopy structure, rooting depth, stomatal regulation, and water-use strategies, which influence aerodynamic coupling, transpiration, and plant water status under wind stress (Kim et al., 2014; Holdo and Nippert, 2015). For example, forests generally have greater canopy buffering and deeper access to soil water, which may prolong physiological responses to wind, whereas grasslands, with shorter and more open canopies and greater reliance on near-surface soil water, may respond more rapidly to wind-induced water loss. Additionally, soil organic carbon emerged as a key factor due to its influence on soil structure, water retention, and nutrient availability. Higher organic carbon enhances resistance to wind-induced moisture loss, thereby moderating temporal effects of wind (Feldman et al., 2018). Elevation, as the main topographic factor, reflects combined gradients in temperature, precipitation, and atmospheric pressure that shape wind-vegetation interactions.

Several limitations should be acknowledged. First, the analysis is based on gridded datasets, which may not fully capture local-scale heterogeneity or extreme events. Second, the statistical methods used to characterize lagged and cumulative effects do not explicitly resolve the underlying physiological processes, which may limit the interpretation of causal relationships. Moreover, testing multiple lagged and cumulative timescale combinations and selecting the combination with the highest significant R 2 may introduce selection uncertainty and increase the likelihood of chance associations. Thus, the identified optimal timescales should be interpreted with caution as statistically optimal timescales within the predefined search range rather than definitive physiological response times. Third, potential interactions among multiple climatic drivers were not fully explored, and vegetation responses to climate drivers may be nonlinear. In particular, the lagged and cumulative analyses primarily focused on the relationship between wind speed and GPP, whereas GPP is also influenced by other climatic factors, such as temperature and precipitation, which may interact with or modify wind effects. Therefore, the observed lagged and cumulative patterns should be interpreted as associations rather than definitive causal effects of wind on vegetation productivity. Future studies should integrate process-based models, field observations, and experimental approaches to better constrain the mechanisms underlying wind-vegetation interactions.

This study presents a global analysis of wind-induced temporal effects on vegetation productivity and their drivers over the period 2000–2024. We found that wind was significantly associated with lagged and cumulative variations in vegetation productivity in 37% of global ecosystems. Wind-induced lagged and cumulative effects primarily occurred at short time scales (≤1 month), with a global mean lagged and cumulative time of wind of 0.74 ± 0.96 and 0.89 ± 0.99 months, respectively. Moreover, climate factors are the primary drivers of these temporal dynamics (∼31%), followed by soil (∼27%), topography (∼23%), and other related factors. The lagged time generally increases with precipitation, while the cumulative time increases with air temperature across most vegetation types. These results underscore the role of wind in vegetation productivity and the importance of accounting for temporal effects in vegetation-wind analyses.