Global climate has been in a state of continuous warming for nearly a century. The current rate of temperature increase is approximately twice that of the previous century (Karl et al., 2015), and this increase is most pronounced at high elevations and latitudes (Peñuelas et al., 2013). Several studies have focused on the effects of climate change on plant diversity in different regions of the world (Boutin et al., 2017; Harrison et al., 2020). The species richness of vascular plants has also increased with the rise in temperature and nitrogen deposition, resulting in notable species-composition shifts (Boutin et al., 2017). In the Columbia River Gorge National Scenic Area, species richness, annual average temperature, and relative humidity were found to be significantly and positively related to each other (Matos et al., 2017). The higher the plant species diversity, the lower the impact of climate change (Li et al., 2018). Globally, regions with warm and wet climates support more species than those with cold or arid climates. This broad-scale climatic influence outweighs any other contributor to plant species diversity (Kreft & Jetz, 2007; Harrison et al., 2020). Reportedly, taxonomic diversity increases with increasing rainfall or varies with increasing productivity despite a slight decrease in temperature (Kreft & Jetz, 2007; Harrison et al., 2020). The relationship between woody species composition and climate is highly consistent across spatial scales and organizational levels (Kreft & Jetz, 2007; Harrison et al., 2020). Based on a very large dataset of six million trees in more than 180,000 field plots in central Africa, researchers have shown that sensitivity to climate change is the highest in endemic species-dominated forests and the driest forests (Réjou-Méchain et al., 2021). Further, recent studies in West Africa have shown that dry tropical forests exhibit larger functional changes compared with moist forests in response to long-term drought (Aguirre-Gutiérrez et al., 2019) and are likely to be more sensitive to global warming (Sullivan et al., 2020). In another study conducted in the Amazon, researchers found a peak in phylogenetic diversity at an intermediate level of precipitation (Neves et al., 2020). Conversely, forests dominated by widespread tree taxa adapted to anthropogenic pressures show relatively low sensitivity to climate change (Réjou-Méchain et al., 2021). Based on model predictions (Réjou-Méchain et al., 2021), undisturbed semi-deciduous and transitional forests appear phylogenetically more diverse than evergreen forests and demonstrate less sensitivity to climate change. However, these in-depth studies have mainly focused on the effects of climate change on plant species diversity in certain regions. Notably, an overall understanding of the spatial patterns of species diversity across vegetation types on a large scale remains lacking.

The measurement of species diversity on a large spatial scale is expected to be more time- and labor-intensive and expensive than on a small scale. With the use of remote sensing, it is now possible to monitor species diversity in large areas over a short period of time (Rocchini, 2007; Madonsela et al., 2017). Of the many different spectral vegetation indices that serve as proxy measures of species diversity, the coefficient of variation in the Normalized Difference Vegetation Index (CV-NDVI), which indicates the variation in spectral species within a plot, i.e., alpha diversity, is most widely used (Peng et al., 2019).

Forests are more appropriate for observing the effects of climate change than other ecosystems because trees have long growth stages and are less affected by occasional, short-term, or intravariable climatic fluctuations (Engler et al., 2009; Zwiener et al., 2018). Spectral diversity indices extracted from remote sensing imagery are particularly useful for predicting forest species diversity because the size of a tree crown usually matches well with the pixel size of satellite images. Furthermore, the use of protected areas (PAs) in this type of analysis can minimize non-climate anthropogenic impacts on plant diversity. An examination of the spatial distribution of plant species can help us to understand the mechanism of climate change impacting plant diversity and provide a reference for biodiversity conservation in world forests. In this study, we used spectral diversity (CV-NDVI) to evaluate plant species diversity. Based on the results of previous studies, we hypothesized the following: 1) species diversity could increase due to a rise in global temperature, associated with increased productivity; 2) increased plant diversity would produce strong neighboring effects, and plant–plant competition could become severe; 3) stronger neighboring interactions and plant–plant competition could lead to a longer spatial distance between plants, and the clustering community would become diffused; and 4) the changes in spatial distance and neighboring interactions could produce a feedback effect on species diversity ( Figure 1 ). Using remote sensing techniques and spatial analysis, we tested our hypotheses based on the spectral diversity of vegetation in seven protected forests in China.

This study is the first attempt to link climate change, neighboring interactions, and spatial distance to species richness and evenness. Using high-accuracy modern machine learning models, we identified the key influencing factors that contributed to the variation in spectral diversity on a large scale using 27 climatic variables. Furthermore, we evaluated the contribution of climate, neighboring interactions, and spatial distance to the variation in spectral CV across forest types.

Consistent with previous studies (Zhang et al., 2017), our results indicated that spectral diversity has increased with climate warming. A recent study found that species richness increased on mountain summits in boreal-temperate forests in Europe (Pauli et al., 2012). Over 30 years of succession, species richness and phylogenetic diversity of plantation trees have increased monotonically (Yu et al., 2019). An increase in spectral species richness is closely associated with an increase in the NDVI, effective cumulative temperature, and seasonal variation in moisture availability (Harrison et al., 2015). In heterogeneous grasslands in California (USA), livestock grazing, fire, succession, nitrogen deposition, and exotic species did not contribute to fluctuations in plant diversity (Harrison et al., 2015). In this study, monthly cumulative temperature, rather than annual average temperature, contributed the most to the increase in spectral CV from 1990 to 2020.

We found that neighboring interactions and plant–plant spatial distance increased with increasing species diversity, presumably due to an increase in tree productivity and tree abundance resulting from ecological complementarity. A previously published global meta-analysis demonstrated that diversity effects are prevalent in the most productive environments, whereas abundance effects became dominant under the most limiting conditions (Madrigal-González et al., 2020). Therefore, a higher abundance will definitely affect plant–plant interactions (i.e., neighboring interactions), and the consequences of this may be either positive or negative, depending on species traits, economics, and environmental conditions (Madrigal-González et al., 2020). A meta-analysis showed that larger positive effects favored sapling emergence and survival, whereas smaller negative effects act on plant growth and density (Gómez-Aparicio, 2009). The life form of the interacting species largely influences the outcome of the interaction; for example, herbaceous plants have strong negative effects, especially on other herbaceous species, whereas shrubs have strong supportive effects, especially on trees (Gómez-Aparicio, 2009). Semiarid and tropical ecosystems, but particularly temperate ecosystems, show more positive neighbor effects than wetlands (Gómez-Aparicio, 2009). Biotic interactions are thought to be relatively more important in shaping biodiversity at tropical than at temperate latitudes (Roslin et al., 2017). Ignoring biotic interactions affects the estimate of climatic niche limits that determine the responses of plant diversity to climate change in models (Newbold et al., 2020). The results of the current study underscore the need to include biotic interactions in climate change modeling.

Neighboring species may demonstrate strong positive or negative correlations. However, when both types of correlation exist, the total neighboring interaction in the region may become null or weak, given that positive and negative values counteract each other (Dray et al., 2012). Weak or null values of neighboring interactions may result when both positive and negative correlations shape species spatial distributions (Wagner, 2013; Biswas et al., 2016). In the present study, warming on a regional scale increased plant species richness (spectral CV), creating a stronger spatial correlation between neighboring species, whereas competition on a plant–plant scale created a negative spatial correlation (Biswas et al., 2017). Although neighboring interactions were weak, they still highly contributed to variation in plant species diversity (spectral CV) in our study.

Considering these results, it is reasonable to expect that stronger neighboring interactions will enhance plant–plant competition, thereby producing a negative effect on intraspecific associations. Conclusively, only species distributed over large distances will survive. Therefore, climate warming leads to a large plant–plant distance. Hypotheses 2 and 3 were confirmed by our data. Specifically, we found that the spatial distance between the same spectral species increased with increasing spectral CV under climate change. The spatial patterns of communities are shaped by environmental filtering and biological competition. Environmental filtering produces a spatially aggregated pattern, whereas competition produces a spatially segregated pattern (He & Biswas, 2019). Possibly, intraspecies competition played a stronger role than environmental filtering in structuring the communities studied from 1990 to 2020, likely because the soil properties, landforms, and slope remained unchanged during this period, whereas plant diversity (spectral CV) increased. To a certain extent, higher plant abundance and species numbers might have confounded the positive effects of climate warming at a local scale. Moreover, our results confirmed that the contribution of plant–plant distance to species diversity was higher than that of climate variables for all forest types (hypothesis 1). Consistent with these findings, a previously published field experiment has shown that biotic interactions have a strong effect on plant fitness and eventually override the effects of climate (Tomiolo et al., 2015).

The patterns of spectral metrics in response to climate change were the same as those in the field survey. Consequently, we determined that spectral metrics are reliable proxy measures of plant species parameters. Several dissimilarity measurements have been introduced to quantify the overall heterogeneity in plant species composition on a few or multiple sites. However, pairwise dissimilarities do not account for patterns of co-occurrence at more than two sites (Baselga, 2013). Consequently, the average of the pairwise dissimilarities may not accurately reflect the overall compositional heterogeneity at multiple sites (Baselga, 2013). Within a highly diverse plant community, many plant species lived together or were associated with higher stem abundance. In remote sensing images, the former demonstrates a cluster with high heterogeneity, while the latter demonstrates similar neighboring pixels. A dataset based on remote sensing is reliable for analyzing both the spectral diversity and spatial patterns of plant species in forests.

In the current study, we found that geophysical factors including soil pH and clay content tended to have higher effects on tree species diversity. Although climate is changing rapidly, geophysical factors are relatively fixed and are not likely to change significantly over the timescales considered in this analysis. Consequently, geophysical variables were not included in the models. In addition, evolutionary and biogeographic histories, including past diversification processes and environmental changes, influence the distribution of tree diversity, and such factors should be considered in future studies on the spatial patterns of plant diversity.

Our results indicated that climate warming has increased species diversity, which in turn has increased neighboring interactions, ultimately leading to a longer plant–plant distance. However, we found that climate change contributed less to species diversity than neighboring interactions and spatial distance. Conclusively, we did not detect a significant change in the overall neighboring correlation on a regional scale under climate warming, although we did observe an increase in the spatial distance across spectral species. In future studies, the relevant biotic and abiotic factors should be quantified, and an assessment of the relative contribution of abiotic factors to the spatial pattern of species diversity should be performed. The general effects of neighboring interactions on plant diversity should be considered at the global scale, considering all vegetation types under the conditions of ongoing climate change.

The original contributions presented in the study are included in the article/ Supplementary Material . Further inquiries can be directed to the corresponding author.

YP devised the project and developed the research questions and study design. NP and JX processed and analyzed data. All authors contributed to the manuscript writing and editing. All authors contributed to the article and approved the submitted version.